AP Master

Astrophysical Plasmas
Steven J. Schwartz, Christopher J. Owen1 , and David Burgess

Astronomy Unit, Queen Mary, University of London
London E1 4NS, U.K.
19 November 2002; last revision 4 January 2004
1 now at Mullard Space Science Laboratory, Holmbury St. Mary, Surrey, RH5 6NT
2
Contents
Preface 7
1 Introduction and Overview 9
2 Some Basic Concepts 11

2.1 What is a Plasma? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Plasmas in the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Different ways of describing plasmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.2 Conservation of Charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.3 Lorentz Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.4 Magnetohydrodynamics (MHD) . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.5 Kinetic Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.6 Frame transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Some Basic Plasma Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.1 Plasma oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.2 Charge Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5.3 Collisions, Mean Free Path, and Collisionless Plasmas . . . . . . . . . . . . . . 17
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Individual Particle Motion 21

3.1 Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Motion in uniform, static magnetic field . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Motion in uniform, static magnetic and electric fields . . . . . . . . . . . . . . . . . . . 23
3.3.1 Qualitative Treatement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.2 Mathematical Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4 Particle Drifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4.1 Gravitational Drifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4.2 Curvature Drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4.3 Gradient Drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.5 Useful Parameters for Describing Particle Motion . . . . . . . . . . . . . . . . . . . . . 29
3.5.1 Pitch Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5.2 Guiding Centre . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5.3 Magnetic Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.6 Motion in Static, Non-uniform Magnetic Field: ∇B k B . . . . . . . . . . . . . . . . . . 30
3.7 Conservation of Magnetic Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.8 Comment on Relativistic Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.9 Applications of Particle Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3
4 CONTENTS
3.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4 MagnetoHydroDynamics - MHD 35
4.1 One-fluid MHD Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1.1 The Induction Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.1.2 Ideal MHD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Magnetic Field Behaviour in MHD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.1 ∇ × (V × B) Dominant - Convection . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.2 µ01σ ∇2 B Dominant - Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.3 The Magnetic Reynold’s Number . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Flux Freezing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.1 Example - Field Line Draping . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4 The Plasma Cell Model of Astrophysical Plasmas . . . . . . . . . . . . . . . . . . . . . 41
4.5 Electromagnetic forces in MHD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.6 MHD Waves, Equilibria and Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.6.1 MHD waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.6.2 MHD Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6.3 MHD Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.7 Dynamos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.7.1 The Dynamo Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.7.2 Qualitative Dynamo Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.7.3 Mean Field Kinematic Dynamos . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.7.4 α − ω Solar Dynamo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.7.5 Astrophysical Dynamos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5 The Solar Wind 59

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.2 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.3 Why is there a solar wind? - A simple model . . . . . . . . . . . . . . . . . . . . . . . 60
5.3.1 Static Atmosphere: V (r) = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.3.2 Solar Wind Solutions: V (r) 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.3.3 The Interplanetary Magnetic Field (IMF) . . . . . . . . . . . . . . . . . . . . . 65
5.4 The Real Solar Wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.4.1 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.4.2 Other Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6 Magnetic Reconnection 71
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.2 Magnetic Annihilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.2.1 Static Annihilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.2.2 Dynamic Annihilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.3 Models of Magnetic Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.3.1 Examples of Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.3.2 Sweet-Parker Model of Reconnection . . . . . . . . . . . . . . . . . . . . . . . 78
6.3.3 Petschek Reconnection Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.4 Evidence for Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
CONTENTS 5
7 Shocks and Discontinuities 85

7.1 Shocks: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.2 Shocks Without Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.3 Shock Conservation Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.3.1 The Exactly Parallel Shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.3.2 The Exactly Perpendicular Shock . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.3.3 Oblique Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.4 Shock Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.4.1 Real Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.4.2 Different Parameters Make Different Shocks . . . . . . . . . . . . . . . . . . . 96
7.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
8 Multi-Fluid Models 101

8.1 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
8.2 Generalised Ohm’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
8.3 Waves in Cold Two-fluid Plasmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8.3.1 Two-stream Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8.3.2 Electron Whistler Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
8.3.3 Other Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
A Useful Mathematical Results and Physical Constants 111

A.1 Physical Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
A.2 Vector Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
A.3 Vector Operators in Various Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . 112
A.3.1 Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
A.3.2 Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
A.3.3 Spherical Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
B The Mathematics of Waves 115

B.1 Fourier Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
B.2 Waves in Space and Time: 1D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
B.2.1 Phase Speeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
B.2.2 Group Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
B.2.3 Waves in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Index
6 CONTENTS
Preface
This book accompanies the MSc course by the same name offered as part of the Queen Mary MSc in
Astrophysics. The course consists of 12 two-hour lectures. The material is aimed at a broad audience
and, while somewhat theoretical in nature, tries to familiarise students with fundamental plasma pro-
cesses and their astrophysical applications; some undergraduate mathematics, notably vector calculus,
and physics (mechanics, fluids, and electromagnetism) are required, though these are reviewed where
they are needed.
The course has evolved over the years from an original one taught by SJS, then taken over by DB,
then CJO, before returning to SJS and extended from 6 to 12 lectures. The text and problems originated
from printed notes devised by DB which were developed further by CJO during his tenure. SJS is
responsible for the additional material and for the present re-drafting of the text and production via
LATEX into an online version. Many figures have become separated from their sources; others have been
taken from several good web-sites (e.g., outreach pages of NASA), courses at other universities, the
books listed in subsequent pages and/or re-drawn by us. The lack of proper source recognition in this
first attempt at an online text is unforgiveable and we apologise most profusely.
In 12 weeks one cannot possibly do justice to the physics which governs 50-95% of the universe (de-
pending on your belief in dark matter!). Perhaps the most serious crime is the non-relativistic treatment
of particles, fields, and flows. While this excludes topics such as radiation from plasmas and relativistic
jets, most of the important phenomena are well described, at least qualitatively, by this non-relativistic
treatment; in many cases the net effect is merely to replace the mass by the relavististic mass. The
simplification of most of the derivation allows more topics to be introduced, and enables a rather short
course to touch on nearly all topics. Other notable omissions include collision theory and the rich and
intriguing area of kinetic plasma astrophysics.
I am indebted to my colleagues for their efforts in drafting the early attempts at a text, and accept
full responsibilities for the many errors which will pervade this first translation and enlargement. Many
students over the years have posed questions or comments which have contributed to the evolution of
the course.
I hope this text succeeds in communicating at least some of the fascination the plasma state holds for
astrophysicists. Plasmas are everywhere, and the consequences of electromagnetic effects on particles
and flows are omitted at your peril. Perhaps you will gain the insight to pose my favorite and infamous
question “But what about the magnetic field ?” at the next seminar you attend and then sit back and
watch the speaker’s ensuing angst as they struggle to figure out if you’ve found some fatal flaw!
SJS
August 2002
7
8 CONTENTS
Chapter 1
Introduction and Overview
This text is a brief introduction to basic plasma physics and its applications to astrophysics. Due to the
interests of the authors, there is particular emphasis on solar system plasma physics, although many of
the concepts are readily applicable to the wider field. Plasmas are electrified gases in which the atoms
have become dissociated into negatively charged electrons and positively charged ions. It is estimated
that 99% of the (visible) universe is in the plasma state. In that sense, it is the most important of the 4
states of matter.
SI units will be used throughout, but be warned that much of astrophysical literature is written even
now in cgs. This is particularly troublesome for plasma physics because of the frequent appearance of
constants like ε0 and µ0 (in SI units) which replace the fundamental constant c and geometric factor 4π
in the cgs system.
Because of the nature of plasma physics, this text is somewhat theoretical in nature. The mathematics
will be kept simple, but some vector calculus will be used. However, most of the concepts do not rely
on this level of mathematical ability. Vectors are denoted in boldface type, e.g., B, while the magnitude
of a vector will be given in normal italic maths type, e.g., B ≡ |B|. Unit vectors are denoted by a carat,
e.g., b̂.
Additionally, we will need draw on various concepts from general physics, including Newton’s Laws
of motion, some thermodynamics, fluid dynamics and electrodynamics. It is this interplay between the
various (classical) branches of physics that makes plasma physics so interesting.
Other texts abound, and may suit your tastes or interests or otherwise cover material not presented
here. If you want one (affordable) book on Plasma Physics then consider Dendy. If you want one book
about Space Plasma Physics then try Parks (but it costs about £40) or Kivelson and Russell (∼ £30).
Many of the books which have titles such as “Plasma Astrophysics” are, in fact, quite specialized and at
an advanced level, and so are only recommended if you wish to study beyond the level we will address
here. The following is an out-of-date sampling of good books:
• M.G. Kivelson and C.T. Russell (Editors), Introduction to Space Physics, Cambridge University
Press, 1995. Available in paperback.
• G.K. Parks, Physics of Space Plasmas, Addison-Wesley 1991. Quite good. Covers a lot of mate-
rial, but misses details sometimes. Only available in hardback.
• R.O. Dendy, Plasma Dynamics, Oxford, 1990. Good. Available in paperback.
• P.A. Sturrock, Plasma Physics, Cambridge 1994. Good, but frustratingly in cgs units!. Available
in paperback.
• Chen, Introduction to Plasma physics, Plenum 1974. Good, but oriented mainly to laboratory
plasmas.
9
10 CHAPTER 1. INTRODUCTION AND OVERVIEW
• Cairns, Plasma Physics, Blackie (1985) (avail. in paper back) A bit brief in places, but fairly
comprehensive and readable.
• Boyd and Sanderson, Plasma Dynamics, Nelson (sadly out of print, cgs, but very good)
• Clemmow and Dougherty, Electrodynamics of Particles and Plasmas, Addison Wesley 1969 (very
thorough and mathematical treatment)
• Nicholson, Introduction to Plasma Theory, Wiley.
• Parker, Cosmical Magnetic Fields, Clarendon Press, 1979. Advanced.
• E.R. Priest, Solar Magnetohydrodynamics, Reidel, 1984. Exactly what the title says, and well-
written by the country’s leading advocate.
• E.R. Priest (editor), Solar System Magnetic Fields, Reidel, 1985. A nice collection of articles
from a summer school and covering a range of topics.
• Schmidt, Physics of High Temperature plasmas, Academic Press, 1966 (covers the right topics,
but a bit old now, and biased to laboratory plasmas)
• Choudhuri, Arnab Rai, The physics of fluids and plasmas: an introduction for astrophysicists,
Cambridge University Press, 1999. (A thorough treatment of everything from neutral fluids and
gas dynamics through turbulence, plasma kinetics, MHD, etc. and aimed at astrophysicists. Suf-
fers only from being in cgs units.)
Chapter 2
Some Basic Concepts
2.1 What is a Plasma?

A plasma is a quasi-neutral gas consisting of positively and negatively charged particles (usually ions
and electrons) which are subject to electric, magnetic and other forces, and which exhibit collective
behaviour.
Ions and electrons may interact via short range atomic forces (during collisions) and via long range
electro-magnetic forces due to currents and charges in the plasma. (Gravitational forces may also be
important in some applications.) The long range nature of the electromagnetic forces means that plasma
can show collective behaviour, for example, oscillations, instabilities, etc.
Plasmas can also contain some neutral particles (which interact with charged particles via colli-
sions or ionization). Examples include the Earth’s ionosphere, upper atmosphere, interstellar medium,
molecular clouds, etc.
The simplest plasma is formed by ionization of atomic hydrogen, forming plasma of equal numbers
of (low mass) electrons and (heavier) protons.
2.2 Plasmas in the Universe

Most of the (visible) universe is in form of plasma: plasmas form wherever temperatures are high
enough, or radiation is strong enough, to ionize atoms. The realm of plasma physics covers a wide
range of parameter space. For example, Figure 2.1 shows various plasma populations as a function of
electron density and temperature. Note that the density scale covers some 30 orders of magnitude, the
temperature scale about 10 orders of magnitude.
Some examples of astrophysical plasmas are listed below:
• Earth’s (and other planets’) ionosphere (above 60 km) and magnetosphere
• Sun’s and other stars’ atmospheres, and winds
• Comets
• Cosmic Rays (galactic and extra-galactic energetic particles)
• Interstellar medium
• Jets in active galaxies - radio jets and emission
• Pulsars and their magnetosphere
11
12 CHAPTER 2. SOME BASIC CONCEPTS
Figure 2.1: The realm of plasma physics.
• Accretion disks around stars
We will be looking at some applications of basic plasma physics to astrophysical plasmas. However,
we will concentrate on examples from our “local” corner of the universe! This is because: The Sun is
our nearest star and we can study it in great detail; and spacecraft have been exploring the local plasma
environment (Earth’s magnetosphere, solar wind, magnetospheres of other planets, comets, etc.) for over
30 years. Spacecraft can return in situ observations of an astrophysical plasma (electromagnetic fields,
plasma density, temperature, kinetic particle distributions, flows, waves of all sorts and frequencies) - an
astrophysical plasma laboratory in our own backyard!
2.3 Different ways of describing plasmas

Exact The most “exact” way to specify the state of a plasma is to give the positions and velocities of
all the particles and all the fields at all points in space. For a system of N particles, the particle
description would be a phase space of 6N dimensions (each position and velocity has 3 compo-
nents). (Phase space is the combination of configuration space, i.e., ordinary position, and velocity
space.) However, N is too large to use this description: ∼ 1015 in a 1km cube in the interplanetary
medium.
Distribution Function The statistical approach is to use a distribution function as a function of position,
velocity: f (x, v) in a 6 dimensional phase space ( f is also in general a function of time). The
number of particles in an elemental volume at position (x, v) in phase space is then given by:
2.4. GOVERNING EQUATIONS 13
f (x, v)d 3 xd 3 v. The theory of the evolution of the distribution function is known as plasma kinetic
theory, and is the best, but most difficult approach to use.
MHD Another (simpler) approach is to treat the plasma as a conducting fluid, and adapt the equations
of fluid dynamics to include the effects of electric and magnetic forces. The resulting theory in
its simplest (one-fluid) form, is called MagnetoHydroDynamics, or MHD. MHD can be derived
from the kinetic theory, in certain limits, but this involves averaging out almost all the interesting
kinetic effects. MHD is applicable to most astrophysical plasmas, but many interesting effects
cannot be adequately described using MHD and require either a multi-fluid approach or kinetic
theory. We will discuss MHD in Chapter 4.
Of course, whatever the description of the plasma matter, we still need to know the electromagnetic
fields as functions of time and space.
2.4 Governing Equations

Most astrophysical plasmas are of low density, and so quantum-mechanical effects can be neglected;
plasmas are dominated by long range forces.
2.4.1 Maxwell’s Equations
∂B
∇×E = − (2.1)
∂t
1 ∂E
∇ × B = µ0 j + (2.2)
c2 ∂t
ρq
∇·E = (2.3)
ε0
∇·B = 0 (2.4)
where µ0 ε0 = 1/c2 , ρq is the charge density, and j is the current density.

Maxwell’s equations can also be given in terms of electric displacement D and magnetic induction
H. In the case of the plasmas we are considering, individual particles contribute to the charge density
and electric currents as though they move in a vacuum, so that D = ε0 E and B = µ0 H.
2.4.2 Conservation of Charge

In the absence of ionization or recombination, then one also has:
∂ρq
∇·j+ =0 (2.5)
∂t
2.4.3 Lorentz Force

Force on single particle of charge q and velocity v:
F = q(E + v × B) (2.6)
For an extended medium there is a force density of ρq E + j × B

2.4.4 Magnetohydrodynamics (MHD)

The derivation of the MHD equations will be covered in Chapter 4, but reduce to straightforward fluid
equations:
∂ρ
+ ∇ · (ρV) = 0 (2.7)
∂t
∂
!
1
ρ + V · ∇ V = −∇p + (∇ × B) × B (2.8)
∂t µ0
∂
!
+V·∇ pρ−γ

= 0 (2.9)
∂t
∂B 1 2
= ∇ × (V × B) + ∇ B (2.10)
∂t µ0 σ
E + V × B = j/σ (2.11)
where σ is the electrical conductivity.
2.4.5 Kinetic Theory

The evolution of the distribution function fα (x, v,t) for particles of species α in space and time in the
absence of collisiones is called the Vlasov Equation:
∂ fα qα ∂fα
+ v · ∇fα + (E + v × B) · =0 (2.12)
∂t mα ∂v
coupled with Maxwell’s Equations (Section 2.4.1) in which the charge and current densities are deter-
mined from the particle distributions as
ZZZ
ρq = ∑ qα f d 3 v (2.13)
α
V
ZZZ
j = ∑ qα v f d 3 v (2.14)
α
V
This multi-dimensional coupled system of integro-partial differential equations is obviously not easy to
solve, but does display the scope for a rich variety and complexity of plasma phenomena. This approach
is largely beyond the scope of the present text.
2.4.6 Frame transformations

The electric and magnetic fields E and B are frame dependent quantities and so in order to transform
between frames the appropriate (relativistic) Lorentz transformation has to be used.
Consider a ‘laboratory’ coordinate frame in which the fields are E and B. Consider now the fields
E0 and B0 in another frame which is moving relative to the first at a velocity u. Then the fields in the
moving frame are given by:
E0k = Ek (2.15)
B0k = Bk (2.16)
2.5. SOME BASIC PLASMA PHENOMENA 15
E⊥ + u × B
E0⊥ = q (2.17)
1 − (u2 /c2 )
u×E
B⊥ −
c2
B0⊥ = q (2.18)
1 − (u2 /c2 )
where k denotes the component parallel to the transformation velocity u and ⊥ the components perpen-
dicular to u. Often u2 c2 , and over sufficiently large scales the natural electrostatic fields are small,
and so in this case: E0 ≈ E + u × B and B0 ≈ B (i.e., magnetic fields are approximately absolute, but
electric fields are not). In the simple case where E, B, and u are mutually perpendicular, Equations 2.15
and 2.16 show that this property is retained in the primed frame. Conversely, while it is possible to trans-
form away the perpendicular electric field, say, by choosing u such that u × B = −E⊥ , any component
of E parallel to this u will remain.
2.5 Some Basic Plasma Phenomena

As a simple introduction to an example of plasma collective behaviour we will consider how a plasma
tries to stay charge neutral, over which scales it succeeds, and the associated characteristic frequency.
2.5.1 Plasma oscillations

Consider a plasma with equal numbers of positively and negatively charged particles. At sufficiently
large distances this produces a vanishingly small electric field, since all the contributions from positive
and negative charges cancel. The compensation is only statistical since all the particles will have random
thermal velocities. On small scales one expects there to be local breakdowns in charge neutrality.
But, since plasmas consist of positive and negative charges, and opposite charges attract, then any
departure from charge neutrality leads to a restoring force back towards charge neutrality. This restoring
force leads to a natural oscillation of the plasma, which are called plasma oscillations and occur at a
frequency called the plasma frequency.
In a plasma consisting of electrons and ions, the electrons move much more rapidly than the ions,
so they are mostly responsible for this oscillation. In this case consider the following simple model,
sketched in Figure 2.2: the ions are fixed, and the electrons have no thermal motion. A single, planar
slab (thickness L) of electrons is displaced a distance ∆x in the x direction (we assume that ∆x L). If
the particle number density is uniform ne = ni = n0 , then this displacement produces two charge sheets
(of opposite sign) with charge per unit area of σ = en0 ∆x, separated by distance L. This is like a parallel
plate capacitor, with a resulting electric field E = σ/ε0 = en0 ∆x/ε0 (from Poisson’s equation).
The electron slab has mass per unit area of me n0 L, charge per unit area of −en0 L and is subject to
the electric field E. Thus equation of motion for the displace ∆x of the electron sheet is (from F = ma):
d 2 ∆x
(−eno L)E = me n0 L (2.19)
dt 2
or
d 2 ∆x
2
= −ω2pe ∆x (2.20)
dt
which is equation for simple harmonic motion at an (angular) frequency ω2pe . The oscillations are called
electron plasma oscillations and the frequency is one of the most important characteristic frequencies of
a plasma.
Figure 2.2: The displacement of a slab of electrons establishes an electric field which tries to return the
electrons, leading to plasma oscillations.
n0 e2
Electron Plasma Frequency ω2pe = (2.21)
me ε0
Plasma oscillations are a direct result of the plasma trying to maintain charge neutrality. They are
independent of the “wavelength” L of the oscillation, unlike, e.g., sound or light waves. One finds that
f pe (kHz) ≡ ω pe /2π ≈ 9 ne (cm−3 ). In the Solar System f pe changes from many Mhz in the corona, to
p
∼ 50kHz at the orbit of the Earth, to < 1kHz far from the Sun.
Plasma oscillations can be driven by the natural thermal motions of the electrons (thermal noise at
the plasma frequency). In this case one can equate the work done by moving the electron sheet through
R e2 n0 x
a distance ∆x by integrating 0∆x eE(x) dx = 0∆x
R
dx with the average energy in thermal agitation in
ε0
the x direction
e2 n0 ∆x2 1
≈ kb T (2.22)
2ε0 2
where T is temperature and kb is Boltzmann constant. Consequently the maximum distance that the
electron sheet can move due to thermal agitation is ∆xmax = λD where:
s
ε0 kb T
Debye Length λD = (2.23)
e2 n0
2.5. SOME BASIC PLASMA PHENOMENA 17
The Debye length is the spatial scale over which charge neutrality is violated by spontaneous fluc-
tuations in plasma. In order for a charged gas to show collective behaviour (i.e. to be considered a
plasma) the Debye length must be much smaller than the spatial scale of the system (i.e. λD L). In
addition, the number of particles, ND , within the Debye “sphere” (i.e., a sphere of radius λD ) must be
large (ND 1).
Debye Number ND = n0 λ3D (2.24)
For interplanetary space, λD is of the order of a few metres, and a few cm in the solar corona; the number
of particles in a Debye sphere, ND ∼ 108 in each case. Some lines of constant Debye length, λD , and
number of particles in the Debye sphere, ND , were plotted in Figure 2.1.
(Note that there is a third condition for a charged gas to be considered a plasma - that the collision
frequency with neutral particles is small compared to the plasma frequency. If this is not satisfied the
gas behaves as if it were a neutral gas.)
2.5.2 Charge Shielding

One can approach the question of the plasma restoring itself to quasi-neutrality from another direction:
namely, by asking how an initially charge neutral plasma responds to a test (fixed) charge, and calculating
the electric potential φ in the plasma as a result of this additional charge. In vacuum the potential of the
charge would just fall off as 1/r, i.e., (φ = q/(4πε0 r) = A/r. However, in a plasma the charge balancing
forces result in a redistribution of the plasma charges, so that the potential has the form:
A
φ= e−r/λD (2.25)
r
Thus the effect of the test charge (i.e., the electrostatic potential it induces) falls off faster than it
would in vacuum. This effect is called charge shielding, and one sees that the test charge is effectively
screened out beyond r ∼ λD . For charge shielding to be effective there must be sufficient particles in a
Debye sphere, i.e., ND 1.
2.5.3 Collisions, Mean Free Path, and Collisionless Plasmas

Neutral particles have quite small collision cross sections (effective size as seen by another particle)
because they only interact via short range forces (electron shells have to overlap). But for charged
particles the Coulomb force is long range, so collisions in a plasma are in reality a lot of small angle
deflections, rather than a few large angle changes.
(What follows are some order of magnitude arguments. . .)
Consider an ionized hydrogen plasma. Electrical and thermal conductivity is effectively controlled
by electrons (because they are so much more mobile - higher thermal speed - than ions), and their
interaction frequency with ions. Suppose a plasma has temperature T , and the ions are effectively at rest
with respect to the electrons. The thermal energy of an electron is 3kb T /2. An electron will be affected
by a neighbouring ion if the Coulomb potential is of the order (or more than) the electron thermal energy.
This defines a Coulomb interaction distance, rC :
q2 3
≈ kb T (2.26)
4πε0 rC 2
And we can define an effective Coulomb cross-section,

!2
q2 1
σC = πrC2 ≈π (2.27)
6πkb ε0 T2
At T = 106 K (typical of the solar wind plasma) this gives σC ∼ 10−22 m2 which is much larger than the
nuclear cross-section of 10−30 m2 . (N.B.: this is an order of magnitude estimation only since we haven’t
taken account of charge shielding!)
Consider a particle A moving at speed v in a gas of particles B which are assumed at rest, with
number density n, and σ is the collision cross-section for A on B. Suppose that after a collision A is sent
off in a random direction.
The collision frequency is given by σvn (= number of particles in a cylinder of cross sectional area
σ and length vt divided by the time t taken to traverse the cylinder) so that the mean time between
collisions is simply 1/(σvn). So the mean distance travelled by A between collisions, referred to as the
mean free path λm f p is:
1
λm f p = (2.28)
σn
Obviously one can increase λm f p by decreasing n or (for Coulomb interactions) increasing T . It is
possible that λm f p becomes greater than the size of the system, and in this case the plasma is collisionless,
i.e., there is no appreciable energy or momentum transfer via collisions. For example, in the case of solar
wind particles they experience only about one collision on their journey from the Sun to the radius of
the Earth’s orbit.
In many astrophysical systems the plasma is collisionless (because of low densities and high tem-
peratures). The motion of particles is then governed by the behaviour of the particles in the large scale
(global) fields.
2.6 Exercises
1. The solar wind at 1AU is a fully ionised proton-electron plasma with density ∼ 5 particles/cm3
and an electron energy ∼ 3eV. Calculate the electron temperature in Kelvin, the Debye length λd ,
Debye number ND , and plasma frequency ω pe . [Be careful with units. We’ve used traditional units
rather than SI ones here. 1eV ≡ 1 electron Volt = 1.6e−19 Joules.]
2. A hydrogen plasma consists of protons and electrons with densities n p = ne = n0 . The electrons
can be considered to be free to move while the ions are considered fixed on a uniform grid due to
their relatively slower thermal motion. A point test charge q is placed in the plasma and establishes
a resulting potential φ, the charged shielded Coulomb potential given in Equation 2.25
A
φ= e−r/λd
r
If the electron population has temperature T , they can be expected to be distributed in a potential
according to the Boltzmann factor ne (r) = n0 e−(−eφ/kb T ) . Assuming |eφ/kb T | 1, expand this
expression for ne (r) to find an expression for the net charge density ρq (r) = en p − ene (r) and use
this to show that the potential φ(r) is consistent with Poisson’s equation ∇2 φ = −ρq /ε0 . (Use the
1 ∂ 2 ∂φ ∂ ∂φ ∂2 φ

1 1
spherical polar form of the Laplacian ∇ φ = 2
2 r + 2 sin θ + 2 2
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂φ2
with spherical symmetry.)
2.6. EXERCISES 19
3. Consider again the plasma oscillation problem posed in Section 2.5.1 and sketched in Figure 2.2
but in which the ions of mass mi are also free to move. Denoting the electron and ion displacements
by ∆xe and ∆xi respectively, find the net surface charge density σ due to these displacements. Write
down the separate equations of motion for the ion and electron slabs and show that they can be
combined to give a single equation of motion for the difference between the two displacements.
Find the frequency of oscillation of this difference. Your answer should reduce in the limit mi
me to ω pe .
Chapter 3
Individual Particle Motion
3.1 Equation of Motion

Individual particle motion can have direct applications in collisionless plasmas, but it also underlies a
lot of the kinetic theory of collective plasma behaviour.
Consider a charged particle, mass m and charge q, moving with velocity v in a given electro-magnetic
field. We will neglect collisions, gravitational and any other forces, and only consider non-relativistic
motion v c. The equation of motion then follows from the Lorentz force:
dv
m = q (E + v × B) (3.1)
dt
From this simple equation a wide range of behaviours is possible depending on the nature of E and B in
space and time.
3.2 Motion in uniform, static magnetic field

(In this section we do not solve the equation of motion directly, but use it to determine the type of
motion.)
Consider case where E = 0 and B is uniform (same everywhere in space) and static (unchanging in
time). The equation of motion becomes:
dv
m = qv × B (3.2)
dt
The force on the particle is perpendicular to v and B, so the magnetic force can do no work on the
particle (rate of work done by a force F in moving something at a velocity v is v · F). So, the kinetic
energy of the particle must be constant.
Check by dotting with v:

dv d 1 2
mv · = v · (qv × B) ⇒ mv = 0 (3.3)
dt dt 2
using v · v = v 2 , and the vector identity A · (A × B) = 0. Hence the kinetic energy W ≡ 21 mv 2 is constant
in time.
Examine the change in v by resolving it into two parts: parallel and perpendicular to B:
v = vk b̂ + v⊥ (3.4)
21
22 CHAPTER 3. INDIVIDUAL PARTICLE MOTION
where vk = v · B/B and b̂ is a unit vector in the direction of B. Then the dotting Equation 3.2 with b̂
yields
dvk
m =0 (3.5)
dt
while the perpendicular part of Equation 3.2 reduces to (recalling b̂ × B = 0)
dv⊥
m = qv × B ≡ qv⊥ × B (3.6)
dt

Thus, vk = constant, and from W = 12 m vk2 + v⊥ 2 = constant, it follows that v changes only in
⊥
direction, not magnitude. The acceleration is centripetal, and the particle will execute uniform circular
motion in the plane perpendicular to the magnetic field direction.
Suppose that the radius of motion is rL . Then the radial equation of motion is (balancing radial
acceleration for constant speed against the perpendicular magnetic force)
mv⊥2
= |qv⊥ B| (3.7)
rL
so the radius of gyration, known as the Larmor radius
mv⊥
Larmor radius rL = (3.8)
|q|B
In one gyration the particle travels distance 2πrL at speed v⊥ , so the angular frequency of gyration
is:
qB
Cyclotron (or Gyro-) Frequency Ωc = (3.9)
m
The cyclotron frequency is one of the most important plasma characteristic frequencies, and is in-
dicative of the field strength and the charge and mass of the particles in the plasma. The cyclotron
frequency is independent of the particle energy. A plasma could have several cyclotron frequencies
corresponding to the different species of the constituent particles, but all particles of the same q/m ra-
tio have the same cyclotron frequency. [For relativistic particle motion, this is no longer true as the
relativistic mass replaces m in Ωc .]
The cyclotron frequency Ωc is usually left as a signed quantity, thereby indicating the sense of
rotation. The sense of gyration is such that the force qv × B points to the centre of gyration. Particles of
opposite signs gyrate in opposite directions. Electrons follow a RH rotation sense with respect to B, as
shown in Figure 3.1.
Particles can have cyclotron motion due to their thermal velocities. For a plasma with electrons and
ions of similar
p temperatures the electron Larmor radius is much smaller than the ion Larmor radius, by
the ratio me /mi . This is important when deciding whether a phenomenon (with a given scale size) is
associated with the electrons or the ions.
Knowing the parallel and perpendicular motion we can now describe the overall motion: In a static,
uniform magnetic field, a charged particle moves along the magnetic field direction at uniform speed
while gyrating around the field at a frequency Ωc , with a radius of gyration rL . This combined motion is
a HELIX of constant pitch along B.
3.3. MOTION IN UNIFORM, STATIC MAGNETIC AND ELECTRIC FIELDS 23
Figure 3.1: Cyclotron motion in a uniform magnetic field for positive and negative charges
3.3 Motion in uniform, static magnetic and electric fields

3.3.1 Qualitative Treatement
In this section we will see the combined effect of a magnetic and electric field. At first we will examine
the type of motion, and then we will show a mathematical solution of the equation of motion.
Consider first the motion of the particle associated with the perpendicular (to B) components of the
electric field, i.e., E⊥ . From the Lorentz transformation (Section 2.4.6) for E and B, we can transfer to
another frame, moving at velocity vD perpendicular to B, in which the magnetic field is unchanged, but
the electric field is given by
E0⊥ = E⊥ + vD × B (3.10)
This means that by a suitable choice of vD we can arrange that E0⊥ is zero, namely
E×B
E × B Drift vD = (3.11)
B2
(Check by substitution and using the vector identities (E × B) × B = (E · B) B − (B · B) E and E · B ≡

0 here since there is only an E⊥ ). We leave the full electric field E in Equation 3.11 since any parallel
component leaves vD unchanged.
In this new frame there is no perpendicular electric field, and so the perpendicular particle motion
is just controlled by the magnetic field (which is unchanged by the transformation), and so it is just
uniform circular motion, as previously discussed.
Now returning to the original frame, the particle motion will be uniform circular motion about the
magnetic field, and a superimposed uniform drift vD which has the initially surprising property of be-
ing perpendicular to both E and B. This is known as the E-cross-B drift velocity, and is sketched in
Figure 3.2. Note that particles of opposite charge drift in the same direction at the same speed, a result
which follows immediately from the fact that in a frame moving with vD the electric field vanishes and
Figure 3.2: E × B drift of particles in steady, uniform electric and magnetic fields
so all particles simply perform cyclotron motion in that frame. The perpendicular motion in the initial
frame describes a CYCLOID.
Qualitatively, the E × B drift can be seen without recourse to frame transformations by considering
the various portions of the particle gyration around the magnetic field. When a q > 0 particle is travelling
in roughly the same direction as E it gets accelerated, increasing its speed. This increasing speed leads
to an increasing radius of gyration (see Equation 3.8) over this portion of the orbit (see Figure 3.2).
But the magnetic field turns the particle until eventually it is moving against the electric field, so it
systematically loses speed and moves in a tighter arc. Over a complete gyration period, the consequence
of larger arcs at the top of each orbit shown and smaller arcs around the bottom of each orbit results in
a net drift. Negatively charged particles circulate in the opposite sense (Figure 3.1) but decrease their
speed when moving along E so perform smaller arcs at the top of each orbit and hence drift in the same
direction as q > 0 particles. That both categories of particle drift at exactly the same rate is easiest to see
via the frame transformation arguments described above.
We now examine the behaviour in detail, to include parallel motion. From our discussion above it
seems that the motion can be simplified by splitting the particle motion into two parts: v = u + vD . Then
the LHS of the equation of motion 3.1 becomes
dv du dvD
m =m +m (3.12)
dt dt dt
But the second term is zero, as vD is constant, and hence, substituting for v and vD , the equation of
motion becomes
du E×B
m = qE + q (u × B) + q 2 × B (3.13)
dt B
Using the same vector identity as before, (E × B) × B = (E · B) B − (B · B) E, this simplifies to
!
du E·B
m = qE + q (u × B) + q − qE
dt B2 (3.14)

≡ q (u × B) + q E · b̂ b̂
3.3. MOTION IN UNIFORM, STATIC MAGNETIC AND ELECTRIC FIELDS 25
where b̂ ≡ B/B is a unit vector in the direction of B. Here the perpendicular components of the E-field
disappears, leaving only the parallel component via E · B. We now split u into components parallel and
perpendicular to B
duk
m = qEk (3.15)
dt
du⊥
m = q (u⊥ × B) (3.16)
dt
Thus, the first equation implies parallel motion consisting of uniform acceleration driven by the
parallel electric field. The second equation, as we have seen already, leads to uniform circular motion
perpendicular to the magnetic field.
We can now return to the complete motion described by
v ≡ uk b̂ + u⊥ + vD (3.17)
and comprised of, respectively: uniform acceleration parallel to B due to any parallel electric field,
uniform gyration perpendicular to B, and a uniform drift perpendicular to both E and B.
3.3.2 Mathematical Treatment

We now tackle the same problem by finding a mathematical solution to the equation of motion
dv
m = q(E + v × B) (3.18)
dt
Splitting the electric field and velocity into components parallel and perpendicular to B one finds
dvk
m = qEk (3.19)
dt
dv⊥
m = q(E⊥ + v⊥ × B) (3.20)
dt
Assuming that the magnetic field is aligned with the ẑ direction, and that the initial parallel position
and speed of the particle are z0 and vk0 , respectively, the solution of the parallel equation of motion is
(vz ≡ vk )
qEk
vk = t + vk0 (3.21)
m
and
qEk
z= t 2 + vk0t + z0 (3.22)
2m
This makes it clear how any parallel electric field accelerates particles along the field. Such parallel
acceleration is an important mechanism for the acceleration of particles. However, since plasma parti-
cles are able to move easily along the field direction, they are able to shield out parallel electric field
perturbations. Such electric fields are thus usually small in astrophysical plasmas.
We now solve the perpendicular motion in the case of static, uniform electric and magnetic field. We
choose the orthogonal geometry: B = (0, 0, B) and E = (E⊥ , 0, 0). The equations of motion for x and y
then become
dvx Ωc E⊥
= + Ωc vy (3.23)
dt B
dvy
= −Ωc vx (3.24)
dt
where we have used Ωc = qB/m. Differentiating the second equation, and using the first, produces
d 2 vy Ω2c E⊥
+ Ω2c vy = − (3.25)
dt 2 B
This is a second order differential equation which can be solved straightforwardly. (Set RHS to zero
to solve for the complementary function, then find the particular integral. The general solution is their
sum.)
E⊥
vy = A cos(Ωct + φ) − (3.26)
B
where A and φ are constants specified through the particle initial conditions. Using this expression for
vy in the equation of motion (3.24 one finds:
vx = −A sin(Ωct + φ) (3.27)
Suppose that the initial velocity (t = 0) is v⊥ = (0, vy0 ), then one finds φ = 0 , and A = vy0 + E⊥ /B,
yielding
!
E⊥
vx = − vy0 + sin(Ωct) (3.28)
B
!
E⊥ E⊥
vy = vy0 + cos(Ωct) − (3.29)
B B
One sees that the particle has a gyration motion (set by initial perpendicular velocity) and a uniform
drift in the ŷ direction, perpendicular to B and E⊥ , which is the E × B drift. Integrating the equations for
vx and vy leads to the equations for the particle position.
3.4 Particle Drifts

We have seen that the addition of an electric field (which results in an additional force on the particle)
to a magnetic field results in a drift of the guiding centre which is perpendicular to the electric force and
the magnetic field. When one considers more general situations, where there is a changing electric field,
or non-uniform magnetic fields, one finds that such particle drifts are a very general feature. There are
guiding centre drifts associated with gravitational force, electric field, polarization (time-varying electric
field), field line curvature, and gradients in the magnetic field strength.
3.4.1 Gravitational Drifts

For example, consider the case where there is a uniform, static magnetic field (E = 0) and the particle is
subject to an external force F, e.g., gravity. We can repeat our analysis by defining an effective electric
field, Ee f f , such that
F ≡ qEe f f (3.30)
3.4. PARTICLE DRIFTS 27
It follows that the motion is again the sum of three parts: acceleration parallel to B, gyration perpendic-
ular to B, and a uniform drift at velocity vF (calculated with Ee f f ), namely
F×B
vF = (3.31)
qB2
since the mathematical equations are identical in form to those presented in Section 3.3.2 The drift is
perpendicular to F and B, and, unlike the E × B drift, depends on the particle charge. So a mixture of
particles with different charges, subject an applied force, can produce so-called drift currents because of
different drift directions for positive and negative charges.
3.4.2 Curvature Drift

Another type of drift is associated with curvature of the magnetic field lines. In this situation we expect
the particle motion to consist of gyration around the field, and motion of the guiding centre along the
curved field line, and is illustrated in Figure 3.3. Relative to the field line, the particle feels an effective
centrifugal force pushing it away from the centre of curvature. This leads to a drift
Fc × B
vc = (3.32)
qB2
via Equation 3.30.

Now, the centrifugal force is given by
Rc
Fc = mvk2 (3.33)
R2c
where Rc is the radius of curvature vector (which points away the centre of curvature). If b̂ · ∇ ≡ ∂/∂s is
the derivative along a field line, i.e., the rate of change as one moves in the direction b̂ ≡ B/B of B, then
∂ B

Rc
= −
R2c ∂s B
1 ∂B B ∂B
= − + (3.34)
B ∂s B2 ∂s
1 B ∂B
= − 2 (B · ∇) B + 2
B B ∂s
We can now substitute this into our expression for the centrifugal force Fc to yield the following result:
mvk2
Curvature Drift vc = [B × (B · ∇) B] (3.35)
qB4
where we have changed the order of the cross product to remove the leading minus sign in Equation 3.34
and noted that B × B = 0. This drift is in the direction Rc × B for positive particles, and in the opposite
direction for negative ones.
Figure 3.3: Particle motion in a curved magnetic field. The curvature is measured by the radius of
curvature, Rc which points away from the centre of curvature. The particle feels an effective centrifugal
force (because there is no explicit centripetal one) and so drifts as if acted upon by Fc .
3.4.3 Gradient Drift
Finally, yet another kind of drift occurs when there is a gradient in the magnetic field in the direction
perpendicular to B. This is called the gradient drift. (In fact, the gradient and curvature drifts usually
occur together since curved field lines are usually associated with perpendicular gradients in the field
strength in nature e.g. in a dipole magnetic field). The origin of this drift is clear if we recall that
the gyroradius of a given particle is inversely proportional to the magnetic field strength. Hence if the
particle is in motion in a region in which the field strength varies across the orbit, the gyroradius also
varies, and leads to a drift which is perpendicular to both B and to ∇B, as illustrated in Figure 3.4.
The mathematical derivation of the gradient drift (usually referred to as the “Grad-B” drift) will not
be shown here. Essentially, the calculation performs a Taylor expansion of the magnetic field strength
and then solves the equation of motion by expansion under the assumption that the Larmor radius is
small by comparison with the length scale over which the field varies. The result, cast back from this
Cartesian calculation to vector notation, is
2
mv⊥
Gradient Drift v∇B = (B × ∇B) (3.36)
2qB3
3.5. USEFUL PARAMETERS FOR DESCRIBING PARTICLE MOTION 29
Figure 3.4: Drift of particles across a gradient in the magnitude of the magnetic field.
3.5 Useful Parameters for Describing Particle Motion

3.5.1 Pitch Angle
The pitch angle is the angle between the velocity vector v and the magnetic field B:
v⊥
α = tan−1 (3.37)
vk
Equivalently vk = v cos α and v⊥ = v sin α. Particles with small pitch angle are called field aligned, and
those with large (i.e., close to 90◦ ) pitch angle are sometimes called gyrating.
3.5.2 Guiding Centre

Particle motion is often decomposed into drifts plus gyration, because often the gyration is on a scale
much smaller than the physical system and/or much more rapid than the phenomena of interest. We talk
about the motion of the average centre of gyration, known as the “guiding centre.” For “slow” changes
in the field the motion of the guiding centre is continuous. One way to calculate the guiding centre
position is to average the particle position over one cyclotron period (i.e., the time taken to complete one
gyration).
3.5.3 Magnetic Moment

Gyration around the magnetic field leads to a current. Consider a charge q circulating with a frequency
Ωc /2π. The equivalent current I is then the charge per second passing a given point, i.e.,
1 qB
I=q (3.38)
2π m
since the total charge passes in one cyclotron period. The magnetic moment of a current loop is the
current carried by the loop times the area of the loop. Thus the corresponding magnetic moment, µm , of
the gyrating particle is given by µm = IπrL2 , or
2
mv⊥ W⊥
Magnetic Moment µm = ≡ (3.39)
2B B
Figure 3.5: Magnetic configuration in which ∇B k B - a “magnetic mirror” geometry.
where W⊥ is the kinetic energy of the perpendicular motion. The magnetic moment µm is also a measure
of the magnetic flux within the particle Larmor radius, a result which follows from its definition and
Equation 3.38 which shows that the field and current are linearly related by constants.
3.6 Motion in Static, Non-uniform Magnetic Field: ∇B k B

Here we examine the motion of a charged particle in the following circumstances:
• a static magnetic field (not time varying)
• a gradient in magnetic field strength in the direction of B, thus the field lines converge, without
any “twist”
• all gradients are “weak” (i.e. any variation in field experienced by particle is slow compared with
cyclotron period)
• zero electric field (therefore, total energy conserved).
This configuration is sketched in Figure 3.5

Assume the magnetic field is mainly in the ẑ direction. Since we are working in the guiding centre
approximation, the particle motion to lowest order is circular, with cyclotron frequency given by the
field strength at the centre of gyration.
∂Bz
The main influence on B is via the variation . However, ∇ · B = 0 implies that there must also be
∂z
∂Bx ∂By
variation via and/or . Here we will use cylindrical polar coordinates (r, φ, z), with unit vectors
∂x ∂y
(r̂, φ̂, ẑ), in which the magnetic field only has components Bz and Br (i.e., no twist). The particle motion
has constant speed (cf: Equation 3.3 which actually holds as long as E = 0 regardless of any spatial
dependence of B).
Expressing ∇ · B = 0 in cylindrical coordinates with axial symmetry
1 ∂ ∂Bz
(rBr ) + =0 (3.40)
r ∂r ∂z
3.6. MOTION IN STATIC, NON-UNIFORM MAGNETIC FIELD: ∇B k B 31
we multiply by r and then integrate with respect to r from r = 0 (which we take to be the position of the
∂Bz
particle’s guiding centre) to r = r. Assuming can be taken as constant over one orbit of the particle,
∂z
this yields
r ∂Bz
Br ≈ − (3.41)
2 ∂z
Thus ∇ · B = 0 necessarily implies that there is a radial component of B if the main (ẑ) component is not
constant in magnitude.
This radial field Br , which is perpendicular to the circular motion, is vital to understanding the
motion of the particle. A particle orbiting around the z-axis at r = rL , encounters a finite Br . So there
will be a Lorentz force associated with Br and vφ . Reference to Figure 3.1 reveals that for all particles,
q
vφ = − v⊥ , or qvφ = −|q|v⊥ , so the force due to the radial field component is
|q|
rL ∂Bz
!
F = −|q|v⊥ φ̂ × − r̂ (3.42)
2 ∂z
Since rL = mv⊥ /|q|B, we see that this force is independent of the particle charge q. Rewriting in
terms of the magnetic moment µm = mv⊥ 2 /2B, this is simply
∂Bz
F = −µm ẑ (3.43)
∂z
This means that the force is opposite to the field strength gradient, i.e., that it tends to push particles out
of stronger field regions.
Using this parallel force (parallel to B, i.e. along ẑ), the equation of motion becomes
dvz ∂B
m ≈ −µm (3.44)
dt ∂z
where we have used B ≈ Bz , i.e., that the field is mainly in the ẑ direction. Multiplying both sides by
dz dv 2 dvz
vz ≡ , and noting that z = 2vz , we find that
dt dt dt
!
d 1 dWk dB
mvz2 ≡ = −µm (3.45)
dt 2 dt dt
where the left hand side is just the rate of change of parallel energy (Wk ), and the right hand side is the
rate of change of field strength as seen by the particle.
However, since E = 0, particle energy W = Wk +W⊥ is constant, i.e.
dW⊥ dWk dB W⊥ dB
=− = +µm = (3.46)
dt dt dt B dt
where we have used µm = W⊥ /B. Therefore

!
1 dW⊥ W⊥ dB d W⊥ dµm
− 2
≡ = =0 (3.47)
B dt B dt dt B dt
This result implies that the magnetic moment, µm , is constant in time (at least in the guiding centre
approximation which we have used). This invariance of the particle magnetic moment turns out to
be a general property of particle motion for both magnetic fields which change in space (as we have
demonstrated) and also in time, provided that the field as felt by the particle varies slowly by comparison
with its gyro-motion.
Since µm has been shown to be constant the parallel force can be written as
∂(µm B)
Fz = − (3.48)
∂z
This shows that for the parallel motion, µm B behaves like a potential. By analogy, there can be
regions where µm B is high enough to exclude certain particles. The consequence is that some particles
moving towards regions of high B will be excluded, so that their parallel motion will reverse. This is
known as a magnetic mirror, and the above force is known as the mirror force.
Since the total energy W is also an invariant of the particle motion,
Wk = W − µm B (3.49)
so that as B increases, Wk decreases. Since Wk can never become less than zero, the particle ceases its
parallel progress when the parallel energy reduces to zero. At this point the particle has a pitch angle
of 90◦ and all its energy is in the perpendicular motion, W = W⊥ . It does not rest here, however, since
Equation 3.48 ensures that there is a parallel force which repels (“reflects”) it away from the higher field
region; the particle thus reverses its parallel motion. This behaviour is known as “magnetic mirroring.”
Note that particles with similar total energy, but smaller pitch angles have smaller W⊥ and hence smaller
µm . These particles therefore penetrate to regions where B has increased to a greater value before being
mirrored, if indeed they reach fields of sufficient magnitude to mirror them at all.
Consider particles with speed v, and pitch angle α0 at a position where B = B0 . By definition
µm sin2 α sin2 α0
v⊥ = v sin α and therefore = = is a constant for the particles. The mirror point is
W B B0
where sin α = 1, i.e., where
B0
Bm = (3.50)
sin2 α0
If a particle is in a region of space between two high field regions, then the particle may be reflected
at one, travel towards the second, and also reflect there. Thus the particle motion is confined to a certain
region of space, bouncing back and forth between the regions of high field; this process is known as
“magnetic trapping.”
3.7 Conservation of Magnetic Moment

The conservation of magnetic moment is equivalent to the statement that it is a constant (or invariant)
of the motion, just as the total kinetic energy is an invariant (for zero electric field). The concept of
invariants is extremely powerful in describing the possible types of particle motion without having to
solve the detailed equations of motion. We have seen that conservation of magnetic moment is valid
when the fields seen by the particle change “slowly” i.e., slower than the gyration of the particle. This
means that magnetic moment conservation is a so-called “adiabatic invariant.”
It can be shown that whenever the particle motion has a certain periodicity, and provided that the
fields change on a time scale slower than this periodicity, then there is an associated adiabatic invari-
ant. The magnetic moment is the first adiabatic invariant, associated with gyration. For a particle in
a magnetic trap there is a second adiabatic invariant associated with the particle’s bounce period. For
3.8. COMMENT ON RELATIVISTIC MOTION 33
particles in a dipole field there is a third adiabatic invariant associated with the periodic curvature drift
of the particle around the dipole. These invariants allow you to draw many important conclusions about
the particle motion, such as which regions are accessible to the particle, the partition of energy between
the particle’s degrees of freedom, and hence what the properties of particles at a given location might
be, etc., without solving a single equation of motion!
3.8 Comment on Relativistic Motion

So far we have only considered non-relativistic motion, but the exact equation of motion is quite straight-
forward  
d m0 v
q  = q (E + v × B) (3.51)
dt 2 2
1 − (v /c )
A full discussion of the resultant motion can be found in various text books, but we just mention
that for E = 0, one finds gyrational motion as before, but now the gyro-frequency is a function of the
relativistic energy (i.e.,. the gyro-frequency reduces as the particle motion becomes more relativistic)
qB q
Ωc = 1 − (v 2 /c2 ) (3.52)
m0
3.9 Applications of Particle Motion

There are a wide variety of astrophysical environments where conecpts from single particle motion play
a key role. These include
Particle Trapping due to magnetic mirroring in the Earth’s “van Allen” radiation belts and energetic
electrons confined in solar coronal magnetic loops
Particle Transport of energetic particles (galactic cosmic rays, solar energetic particles, ultra-relativistic
particles from extra-galactic sources) is governed by the variety of particle drifts and adiabatic in-
variants discussed in the present chapter. Note that in many regimes, particle transport is not
controlled by collisional, diffusive processes due to the long collision mean free paths, but is
instead controlled by single particle motion. The rate of Coulomb collisions between charged
particles decreases strongly with increasing particle energy.
Particle Acceleration due to parallel electric fields, as found in pulsar magnetospheres and solar flares.
Magnetic mirroring due to either stochastic motion of high field regions, or systematic motion
between converging mirrors in the vicinity of shock waves, leads to Fermi acceleration of particles.
Radiation given off by relavistic gyrating particles (synchrotron radiation) or accelerating particles
(brehmstralung) often provides vital information about the fields and energy content of distant
sources.
3.10 Exercises
1. Calculate the Larmor radius and gyrofrequencies for the following particles
(a) A 10 keV electron moving with a pitch angle of 45◦ with respect to the Earth’s magnetic
field of 30,000nT [nanoTeslas].
(b) A solar wind proton moving at 400 km/s perpendicular to a field of 5nT.
(c) A 1 keV He+ ion in the solar atmosphere near a sunspot where B = 5 × 10−2 T (vk = 0).
2. The magnetic field strength in the Earth’s magnetic equatorial plane is given by
!3
Re
B = B0
r
where B0 = 0.3 × 10−4 T, Re is the Earth radius (∼ 6400km), and r is the geocentric distance.
Derive an expression for the drift period (the time for 1 orbit of the Earth) of the particle under
the influence of the ∇B drift. Evaluate this period for both a proton and an electron of 1keV at a
distance 5Re from the Earth. Compare the answer to (a) the drift induced by the force of gravity
on the same particles and (b) the orbital period of an uncharged particle at the same position. (The
mass of the Earth is 6 × 1024 kg).
3. (a bit harder) A non-relativistic particle of mass m and charge q moves in a steady, axially sym-
metric magnetic field of a sunspot, which is taken to be uniform below the solar surface and falls
off with height, h, above the surface as
H3
B = B0
H 3 + h3
for h > 0, where H is a constant and B0 is the value of the field at and below the surface. [In effect,
the field is roughly dipolar for large h].
(a) Sketch the field lines in this model of a sunspot field

(b) If particles of speed v are produced at the surface with an isotropic distribution (i.e., they
move equally likely in all directions), and conserve their magnetic moments µm , show that
as they move to successively
greater heights h the particles possess a smaller range of pitch
angles, α ≡ arccos vk /v . [Hint: Draw a picture of a spherical shell in (vk , v⊥ ) space, and
consider those particles with pitch angles near 90◦ .]
(c) Hence explain how this process of “magnetic focussing” produces a “beam-like” particle
distribution. How narrow is this beam in angular spread (i.e., pitch angle) at h ∼ 2H?
4. A cosmic ray proton is trapped between 2 moving magnetic mirrors in which the field strength
increases by a factor of 5. It has an initial kinetic energy W = 1keV and v⊥ = vk in the mid-
plane between the two mirrors. Each mirror is moving towards this mid-plane with velocity Vm =
10km/s. Initially the mirrors are separated by a distance L = 1010 km.
(a) Using the invariance of µm find the energy to which the proton will be accelerated before it
escapes the system.
(b) How long will it take to reach that energy?
Hints:
(a) Treat the mirrors as flat pistons and show that vk increases by 2Vm at each bounce.
(b) Compute the number of bounces needed
(c) Accuracy to a factor of 2 is sufficient.
Chapter 4
MagnetoHydroDynamics - MHD
4.1 One-fluid MHD Equations

The magnetohydrodynamic (MHD) equations describe the plasma as a conducting fluid with conduc-
tivity σ1 which experiences electric and magnetic forces. The fluid is specified by a mass density ρ,
a flow velocity V, and a pressure p, which are all functions of space and time. For an electron-proton
plasma the mass density ρ is related to the number density n by ρ = n (mi + me ). Perhaps the most
elegant derivation of the MHD equations, which reveals the many assumptions, is by taking moments
of the Vlasov Equation (see Section 2.4.5), i.e., weighting by successive powers of v and integrating
over all velocity space. See, for example, the treatment in the book by Boyd and Sanderson listed in
the Introduction (p9). Here, we shall simply draw on the ordinary equations of hydrodynamics. The
equation of motion has an additional force per unit volume j × B due to the net magnetic force on the
plasma particles. On fluid scales ( λD ) the plasma is overall charge neutral, so that the net electric
force is neglibly small. The MHD equations represent the conservation of mass, the conservation of
momentum, and some equation of state (i.e., an energy relationship). Thusly we reach:
∂ρ
+ ∇ · (ρV) = 0 (4.1)
∂t
∂
!
ρ + V · ∇ V = −∇p + j × B (4.2)
∂t
∂
!
+V·∇ pρ−γ

= 0 (4.3)
∂t
(4.4)
The operator
∂
!
d
≡ +V·∇ (4.5)
dt ∂t
is called the convective derivative (or total or substantial or Lagrangian derivative). It measures the rate
of change as seen by a parcel of the fluid. This parcel sees any instrinsic time variation, and additionally
sees variations which arise because it is moving (or “convecting”) to other regions where, due to spatial
gradients, the fluid parameters are different.
There are a number of assumptions behind the above equations: The energy equation is the adiabatic
equation of state with γ being the ratio of specific heats (γ = 5/3 for a monatomic gas); there are other
1 Sometimes the resistivity 1/σ is used, and denoted η. However, we shall reserve η for the magnetic diffusivity, which is
1/(µ0 σ)
35
36 CHAPTER 4. MAGNETOHYDRODYNAMICS - MHD
possible choices for an energy equation such as isothermal (γ = 1), etc. Ohmic heating j · E has been
ignored, i.e., assumed to be less important than the plasma’s thermal energy; radiative or other heat
losses are also neglected (or modelled in an ad hoc fashion by playing with the value of γ). We have also
assumed that the pressure is isotropic which is not generally true when the magnetic field can impose
different kinds of motion on the particles parallel and perpendicular to the field direction. Any possible
heat flux has been ignored, and there is no relative flow between the different species in the plasma, so
that the one fluid description is valid. While the conductivity is usually large due to the mobile electrons,
the absence of collisions has the opposite impact on other transport processes such as viscosity, which
we have also therefore neglected.
In addition to the equation for the fluid, one also has Maxwell’s equations for the fields:
∂B
∇×E = − (4.6)
∂t
∇ × B = µ0 j (4.7)
∇·B = 0 (4.8)
where µ0 ε0 = 1/c2 .
Note that the displacement current term in the ∇ × B equation has been dropped because we are
interested in low frequency (gross fluid) behaviour (see Section 2.4.1). Furthermore, the Poisson equa-
tion for ∇ · E has been dropped since we are treating the plasma as a single, charge neutral fluid. These
approximations are discussed later. Chapter 8 relaxes the one-fluid assumption.
While the fluid equations effectively add the mass densities, forces, etc., for all the plasma con-
stituents, the plasma’s conducting properties require some understanding of the differences between
positive and negative species to determine the current density. This results in an Ohm’s Law relating the
electric field to the current density:
j
One Fluid MHD Ohm’s Law E = −V × B + (4.9)
σ
This simply says that in the frame moving with the fluid (see Section 2.4.6) the fluid behaves like a simple
conductor with E0 ∝ j. The term −V × B is called the motional electric field, and is a consequence of the
motion of the fluid and the rule for transforming electric field between frames. Chapter 8 develops the
two fluid approach and derives the generalized Ohm’s Law (Equation 8.17) which includes terms due
to finite electron inertia, so-called Hall effects, thermoelectric fields, and other proesses. The one-fluid
MHD Ohm’s Law is a simplification of this more general result when two-fluid effects are ignored.
4.1.1 The Induction Equation

Combining the fluid and field equations one can find the governing equation for the evolution of the
magnetic field and thereby eliminate the electric field E. Taking the curl of Equation 4.9 gives
1
∇ × E = −∇ × (V × B) + ∇ × j (4.10)
σ
assuming σ is constant. Substituting for j = ∇ × B/µ0 from Ampere’s law and using the law of induction
Equation 4.6 yields
∂B 1
− = −∇ × (V × B) + ∇ × (∇ × B) (4.11)
∂t µ0 σ
4.2. MAGNETIC FIELD BEHAVIOUR IN MHD 37
The double curl can be expanding with the help of the vector identity A.13 to give
∂B 1 1
− = −∇ × (V × B) + ∇ (∇ · B) − ∇2 B (4.12)
∂t µ0 σ µ0 σ
The middle term is zero by Maxwell’s equation for ∇ · B and so the law of induction in a plasma reduces
to
∂B 1 2
MHD Induction Equation = ∇ × (V × B) + ∇ B (4.13)
∂t µ0 σ
This equation, together with the fluid mass, momentum and energy equations 4.1-4.3 form a close
set of equations for the MHD state variables (ρ, V, p, B).
4.1.2 Ideal MHD

The case where the conductivity is high (σ → ∞), such that the electric field is E = −V×B (i.e., motional
electric field only), is known as ideal magnetohydrodynamics.
4.2 Magnetic Field Behaviour in MHD

We now examine the form of the induction equation in two extreme cases, depending on which of the
two terms on the RHS is dominant.
4.2.1 ∇ × (V × B) Dominant - Convection

In this case, corresponding to the infinite conductivity limit dubbed Ideal MHD, the flow and the field
are intimately connected. We shall see that the field lines convect with the flow; conductors do not allow
new fields to penetrate them, and retain any internal fields they possess. Plasma physicists refer to this
as “flux freezing”, which we shall explore more below. The flow in turn responds to the field via the
j × B force.
µ0 σ ∇ B
1 2
4.2.2 Dominant - Diffusion
In this case the induction equation takes the form of a diffusion equation: the field lines diffuse through
the plasma down any field gradients, so as to reduce field gradients. There is no coupling between the
magnetic field and fluid flow.
4.2.3 The Magnetic Reynold’s Number

Now consider the relative sizes of the two RHS terms in the Induction Equation, by using dimensional
arguments. We replace ∇ by 1/L where L is the characteristic scale length, and let V denote the charac-
teristic (i.e. typical) speed. Then the ratio of the convection term to the diffusion term can be expressed
by the dimensionless number
V B/L
Magnetic Reynold’s Number Rm ≡ = µ0 σV L (4.14)
B/µ0 σL2
If Rm is large then convection dominates, and the magnetic field is frozen into the plasma. Else if Rm
is small then diffusion dominates. In the solar system, and in astrophysics generally, Rm is very large,
e.g., 108 in a solar flare, and 1011 in the solar system and planetary magnetospheres. We shall see that
Rm is not large everywhere; thin boundary layers form where Rm ∼ 1 and ideal MHD breaks down. As
in most branches of physics, these boundary layers mediate the global dynamics by controlling the rate
of transport of mass, momentum, and energy through the system.
4.3 Flux Freezing

In 1942 Alfvén showed that the induction equation for ideal MHD is equivalent to the statement that the
“field is frozen into the fluid” or that the “matter of the plasma is fastened to the lines of magnetic force.”
This is an extremely important concept in MHD, since it allows us to study the evolution of the field,
particularly the topology of the field lines, by finding out about the plasma flow. Of course, the principle
also works in the other direction: If we know how the magnetic field lines evolve, then we can deduce
the plasma fluid flow.
To establish this Flux Freezing Law we begin with the Ideal MHD Induction Equation
∂B
= ∇ × (V × B) (4.15)
∂t
and then show that the magnetic flux through a closed loop which moves with the fluid is constant in
time.
The magnetic flux through a closed loop ` is
I
ΦB ≡ B · n̂ dS (4.16)
`
where dS is the area element of any surface which has ` as a perimeter. The quantity ΦB is independent
of the specific surface chosen, as can be proved from ∇ · B = 0 and Gauss’ Theorem. The Flux Freezing
Law we are trying to establish is then expressed as:
dΦB
=0 (4.17)
dt
d
where we use the total derivative to indicate that the time derivative is calculated with respect to fluid
dt
elements moving with the flow.
The quantity ΦB is not locally defined, so we have to follow an explicit calculation for its time
derivative. Figure 4.1 sketches all the necessary components.
Consider a loop of fluid elements ` at two instants in time, t and t + ∆t. The two surfaces S1 and
S2 have `(t) and `(t + ∆t) as perimeters. There is also a “cylinder” S3 generated by the fluid motion
between the two instants of the elements making up `. Let ΦB be the flux enclosed by ` and ΦB1 be the
flux through surface S1 , and similarly for S2 and S3 . The normal vectors to S1 and S2 are chosen to lie
on the same side of the surfaces, relative to the fluid flow.
Then,
ΦB2 (t + ∆t) − ΦB1 (t)
!
dΦB
= lim (4.18)
dt ∆t→0 ∆t
From ∇ · B = 0 the net flux through the three surfaces at any time is zero, so
− ΦB1 (t + ∆t) + ΦB2 (t + ∆t) + ΦB3 (t + ∆t) = 0 (4.19)

4.3. FLUX FREEZING 39
Figure 4.1: Generalized cylinder formed by the motion of a closed line ` “frozen” to the fluid.
(Note negative sign on first term to account for normal pointing inward into the volume, rather than
outward.) We can thus eliminate ΦB2 (t +∆t) , and at the same time use the definition of flux in expressing
ΦB1 and ΦB3  
dΦB 1 ZZ ZZ
= lim (B(t + ∆t) − B(t)) · n̂ dS − B · n̂ dS (4.20)
∆t

dt ∆t→0
S1 S3
The first term is nothing more than
ZZ
∂B
· n̂ dS (4.21)
∂t
S1
Also we can convert the surface integral over S3 to one over S1 by noting that the area element for S3
can be written n̂dS = d` × V∆t, where d` is a line element of the loop of fluid elements. Thus
ZZ I I
B · n̂ dS = B · (d` × V)∆t = (V × B) · d` ∆t (4.22)
S3 `(t) `(t)
Then, by using Stokes Theorem to convert the line integral to a surface integral
ZZ ZZ
B · n̂ dS = ∇ × (V × B) · n̂ dS ∆t (4.23)
S3 S1
And so, finally putting these results into Equation 4.20
dΦB ZZ ∂B
" #
= − ∇ × (V × B) · n̂ dS = 0 (4.24)
dt ∂t
S1
Figure 4.2: Sketch of Solar Wind - Comet interaction illustrating field line draping.
which establishes the Flux Freezing Theorem.

The most important implication of this theorem is that it imposes constraints on the allowable motion
of the plasma. We can, for example, define a magnetic flux tube: Take a closed loop and move it parallel
to the field it intersects. The surface thus created has zero flux through it, and consequently the fluid
elements that form the flux tube at one moment, form the flux tube at all instants. If two fluid elements
are linked by a field line (defined as the intersection of two flux tubes) at one instant, then they are
always so linked. If the conductivity σ, is not infinite (equivalently, Rm is small), then the flux “thaws”
and then slippage of the the field relative to the flow is possible.
4.3.1 Example - Field Line Draping
Consider a flowing plasma with, say, the field lines initially perpendicular to the flow. Suppose that
there is some region where the flow is slowed, or even stopped, by some kind of obstacle for example
as shown in Figure 4.2. Outside of this region the flow continues unabated taking its field lines with it.
But at the obstacle the field lines are slowed, effectively caught up by the obstacle. This has the effect of
stretching out the field lines behind the obstacle, creating a “tail” in the magnetic field behind the region
of slowed flow. In the tail the field reverses direction, so there will be a current sheet at the centre of the
tail. All of the objects visited by spacecraft have been observed to have some kind of magnetic tail: the
Earth, other planets, comets, etc.
4.4. THE PLASMA CELL MODEL OF ASTROPHYSICAL PLASMAS 41
Figure 4.3: Separation of fields and plasmas of different sources by thin boundary layers (current sheets).
4.4 The Plasma Cell Model of Astrophysical Plasmas

A central question is the following: If Rm is large can we neglect the effects of magnetic diffusion
completely?
Large Rm means no cross field plasma mixing: particles are tied to the same field line. Astrophysical
plasmas generally come from different sources, such as stellar winds, or clouds of interstellar gas. But
what happens when plasmas from different sources, carrying different magnetic fields, come in contact?
Figure 4.3 illustrates such a contact between two different plasma regions.
In ideal MHD the lack of plasma mixing means the formation of a thin boundary layer separating
the two different plasma and field systems. The location of the boundary layer is determined by pressure
balance. The boundary layer must also be a current sheet, since the magnetic field changes across it.
Thus we can make the following general model of the plasma universe: Astrophysical plasma sys-
tems become divided into separate cells containing field and plasma from different sources, separated
by current sheets. This is illustrated in Figure 4.4
Observations in the Solar System, where we can actually measure thin layers, indicate that this is an
excellent approximation. As an example we can cite the existence of well defined planetary magneto-
spheres where boundary layers have thickness of order 100 − 1000km in a system with an overall size
of 100,000km.
But there is a problem: The magnetic Reynold’s number (Equation 4.14) was calculated from the
over all length scale L of the system. On the other hand the behaviour of the thin boundary layers
is described by a much smaller Rm . Thus we conclude that diffusion will probably be important at
boundary layers, and, indeed, will be the mechanism of mass and momentum transfer between the two
plasma systems.
4.5 Electromagnetic forces in MHD

Previously we only retained the magnetic force term j × B in the one fluid MHD equations. We now
examine the validity of this assumption.
Generally the equation of motion for the fluid will be of the form
∂
!
ρ + V · ∇ V = −∇p + (e.m. forces) + (viscous forces) + . . . (4.25)
∂t
Figure 4.4: Cartoon illustrating the plasma cell model of the solar system (items NOT to scale!).
Then, treating the plasma as two interpenetrating fluids, one of singly ionised ions and one of elec-
trons (a model we shall explore further in Chapter 8), we can write the electromagnetic forces as
Fe.m. = (ni − ne ) eE + (ni vi − ne ve ) e × B (4.26)

= ρq E + j × B (4.27)
1 ∂E
= ρq E + [(∇ × B) × B] − ε0 ×B (4.28)
µ0 ∂t
where we have used Maxwell’s equations to substitute for j.

We can now compare the relative magnitude of the three terms in the expression for the electromag-
∂
netic force in (4.28), using dimensional arguments, i.e., substituting ∇ ⇒ 1/L, ⇒ 1/T , L/T ⇒ V and
∂t
E ⇒ V B where all the quantities (length, time, velocity) are now the characteristic values for the system
under consideration.
Taking the ratio of the middle to last terms
B2 /Lµ0 T c2
∼ ∼ 1 (4.29)
ε0V B2 /T µ0 ε0 LV V2
∂E
Thus, we can neglect the term, i.e., the displacement current term, provided that the fluid flows are
∂t
much less than the speed of light.
Now consider the ratio of the first to middle terms
ε0 E 2 /L V2
∼ 1 (4.30)
B2 /Lµ0 c2
4.6. MHD WAVES, EQUILIBRIA AND INSTABILITIES 43
where we have used Poisson’s Equation (2.3), i.e., ∇ · E = ρq /ε0 to evaluate ρq . So, again for fluid
velocities much less than the speed of light, the magnetic term dominates over the electrostatic force
term. This completes the justification of using only the magnetic force term in the one fluid MHD
equation of motion (4.2). But we now go further, and find an interesting way of splitting the magnetic
force. Using a vector identity (A.9) we can write
!
1 B2 1
FM = [(∇ × B) × B] = −∇ + (B · ∇) B (4.31)
µ0 2µ0 µ0
We write B = Bb̂, so that b̂ is the unit vector in the direction of the magnetic field. Then, the operator
b̂ · ∇ ≡ ∂/∂s represents the derivative along the direction of the magnetic field where s is a distance
parameter along a field line. Then the last term in (4.31) can be written (e.g., by re-arranging Equa-
tion 3.34)
∂ B2
!
Rc
(B · ∇) B = b̂ − B2 2 (4.32)
∂s 2 Rc
where Rc is the vector from the centre of curvature of the field as sketched in Figure 3.3 We now return
to the expression for FM in (4.31) and split the gradient operator into two parts, one parallel and one
perpendicular to B. The parallel term is just the gradient in the direction of b̂ so that this decomposition
is
∂
∇ = ∇⊥ + b̂ (4.33)
∂s
Thus we can bring things together (cancelling the parallel gradient terms) to arrive at
!
B2 B2 Rc
MHD Magnetic Forces FM = −∇⊥ − (4.34)
2µ0 µ0 R2c
This equation shows that the magnetic force can be resolved into two, conceptually simple, components:
a force perpendicular to the magnetic field which behaves like a pressure (i.e., it is the gradient of a
scalar quantity, B2 /2µ0 ), and a force towards the instantaneous centre of curvature which depends on the
curvature and the field magnitude. This is the physical equivalent of a tension force B2 /µ0 acting along
the field lines. To summarise: Forcing the field lines together results in an opposing perpendicular
pressure force. Trying to bend the field lines results in an opposing tension force. These different kinds
of forces, illustrated in Figure 4.5 lead to different kinds of dynamics.
4.6 MHD Waves, Equilibria and Instabilities

It is from the two kinds of restoring force due to the magnetic field that one can find a number of wave
modes in MHD, and also investigate the equilibria (and instabilities!) of various MHD situations.
4.6.1 MHD waves

From an analysis of the linear wave modes we shall find that, in place of the sound waves associated
with a classical gas, there are three distinct MHD waves:
Alfvén waves In this case the restoring force is due entirely to the tension associated with the field lines;
the wave is transverse and essentially magnetic, with no compression of the matter.
Figure 4.5: Magnetic pressure and tension forces acting on a magnetic flux tube
Fast mode waves The field magnitude and gas pressure both vary. In the fast mode their variation is in
phase, and the waves propagate faster than an Alfvén wave.
Slow mode waves The field magnitude and gas pressure both vary. In the slow mode their variation is
out of phase, and the waves propagates slower than an Alfvén wave.
We will investigate the MHD equations for small amplitude, plane wave solutions. Appendix B
reviews the basic mathematics and conepts related to plane wave analysis which we employ here. Into
the MHD equations we substitute:
B = B0 + B1 (4.35)
V = V0 + V1 ≡ V1 (4.36)
ρ = ρ0 + ρ1 (4.37)
p = p0 + p1 (4.38)
where quantities with subscript “1” represent small, linear perturbations from the values (subscript “0”)
obtained from an equilibrium solution of the MHD equations. Higher order perturbations are ignored.
As noted in (4.36) we will work in a frame where V0 = 0. The equilibrium is assumed to be uniform,
with ρ0 = constant, B0 = constant, and p0 = constant. Then we linearize by substituting into the MHD
equations (see Section 2.4.4), cancel the terms which appear in the equilibrium solutions to obtain
∂ρ1
+ ρ0 ∇ · V1 = 0 (4.39)
∂t
∂V1 1
ρ0 = −∇p1 + (∇ × B1 ) × B0 (4.40)
∂t µ0
∂B1
= ∇ × (V1 × B0 ) (4.41)
∂t
p1 = c2s ρ1 (4.42)
where cs is the sound speed, c2s = γp0 /ρ0 . Now we look for plane wave solutions of the form
Q1 = δQei(k · x − ωt) (4.43)
where Q1 represents one of the fluid perturbation quantities defined above, and where k is the wave
vector (wavelength λ = 2π/|k|), and ω is the angular frequency of the oscillation. Substituting and
equating coefficients of the same exponent will simply give the same result as the following substitution
in the original equations: ∇ ⇒ ik and ∂/∂t ⇒ −iω.
So the equations for the perturbations become
− ωδρ + ρ0 k · δV = 0 (4.44)
1 1
−ρ0 ωδV = −kδp + (k · B0 ) δB − (δB · B0 ) k (4.45)
µ0 µ0
−ωδB = (k · B0 ) δV − (k · δV) B0 (4.46)
δp = c2s δρ (4.47)
This gives us 4 equations containing 4 unknowns. Combining (4.44) and (4.47) gives
ρ0 c2s
δp = k · δV (4.48)
ω
Now substitute (4.48) for δp and (4.46) for δB into (4.45), in order to get an equation for δV
ρ0 c2s
−ρ0 ωδV = − (k · δV) k
ω
(k · δV)

1 (k · B0 )
− (k · B0 ) δV − B0 (4.49)
µ0 ω ω
(k · δV) 2

1 (k · B0 )
+ (B0 · δV) k − B0 k
µ0 ω ω
This can be written in matrix form, AA · δV = 0 , where the matrix A is a function of (k, ω, B0 , ρ0 , p0 ).
In order for this equation to have a non-trivial solution we require detA A = 0. This gives a relationship
between k and ω, which is called the dispersion relation. The corresponding δV is an eigenvector, and
gives the polarization and other characteristics of the wave. While this approach provides the complete
solution to waves in MHD plasmas, we shall content ourselves here to begin by considering in more
detail a few special cases related to choices of the orientation of the wave-vector k with respect to B0 .
Waves propagating parallel to B0

Let B0 = B0 ẑ and k be along the z-direction. First, consider only the z-component of (4.49); both terms
in the square brackets vanish, and one is left with:
ω2 δVz = k2 c2s δVz (4.50)

Figure 4.6: An Alfvén wave propagating parallel to the background magnetic field.
And for δVz 6= 0 this implies ω2 = k2 c2s which is the dispersion relation for sound waves. One finds that
δVx = 0 = δVy and so δV k k, i.e. the waves are longitudinal, with a phase speed of
Sound Wave Dispersion Relation ω/k = ±cs (4.51)
.
Now consider the transverse component of (4.49):
" #
kB0 kB0
− ρ0 ωδV⊥ = 0 − δV⊥ + 0 (4.52)
µ0 ω
so that, for δV⊥ 6= 0,
Alfvén Wave Dispersion Relation ω2 = k2 vA2 (4.53)
where vA2 = B20 /(µ0 ρ0 ) is the square of the Alfvén speed. This is the dispersion relation for Alfvén
waves. The perturbation δV is perpendicular to k and B0 , so that the wave is called a transverse, or shear
wave. The phase speed ω/k = ±vA . For the Alfvén wave, from (4.46)
kB0
δB⊥ = − δV⊥ (4.54)
ω
i.e., the field perturbation and velocity perturbations are parallel to each other. The waves are like waves
on a string, and are due to the field line tension, as sketched in Figure 4.6. Equations (4.44) and (4.48)
tell us that δρ = 0 = δp, so there is no compression of the plasma and the Alfvén wave is thus essentially
magnetic.
Figure 4.7: Fast magnetosonic wave propagating perpendicular (k ⊥ B0 ) (left) and obliquely (right) to
B0 .
Waves propagating perpendicular to B0

From (4.49), k ⊥ B0 gives δVz = 0, and the perpendicular component of (4.49) gives
ρ0 c2s (k · δV⊥ )
− ρ0 ωδV⊥ = − (k · δV⊥ ) k − 0 − B20 k (4.55)
ω µ0 ω
Hence δV⊥ is parallel to k, and thus
ω2 δV⊥ = k2 c2s + vA2 δV⊥

(4.56)
For a non-trivial solution for δV⊥ we therefore reach the dispersion relation
ω2 = k2 c2s + vA2

Fast Magnetosonic Dispersion Relation (4.57)
These are termed fast mode magnetosonic waves. Both B and p show compression (δρ 6= 0, δp 6= 0, δB 6=
0 from (4.44), (4.48) and (4.46)). In this case of perpendicular propagation, the fast mode propagates
with a phase velocity q
ω/k ≡ v f = c2s + vA2 (4.58)
. Figure 4.7 shows the fast magnetosonic mode propagating both perpendicular and obliquely to B0 and
reveals the compression in the field.
General Case: k · B0 = kB0 cos θ

In the general case the wave vector k is oblique to B0 ≡ B0 ẑ, with θ being the angle between the two.
Now there are three waves, essentially because there are three restoring forces: magnetic tension, mag-
netic pressure, and thermal pressure. The dispersion relations for these three modes are
Alfvén Wave ω2 = k2 vA2 cos2 θ ≡ kz2 vA2 (4.59)
The Alfvén wave is purely transverse, does not compress the plasma (δρ = 0 = δp) and carries energy
along the background field (the group velocity ∂ω/∂k = vA ẑ)
" s #
Fast and Slow c2 + vA2 4c2 v 2 cos2 θ
ω2 = k2 s 1± 1 − s A 2 (4.60)
Magnetosonic Waves 2 c2s + vA2
One way to visualise these waves is to imagine dropping a pebble into a plasma “pond” and watching
the resulting ripples. Unlike simple water waves, which propagate at the same speed in all directions
and therefore have circular ripples, the waves in an MHD plasma propagate at different speeds, and
these speeds depend on the direction of propagation. The resulting “ripples” are shown in the Friedrichs
Diagrams in Figure 4.8.
4.6.2 MHD Equilibria

Since there are restoring forces associated with both field line compression and field line bending, there
is the possibility of storing energy in both kinds of field perturbations. This means that, if we consider
an MHD system, we can ask the question: Is the system in equilibrium? Depending on whether the
system is in an energy minimum, or at an energy maximum, we say that it is in a stable or unstable
equilibrium, respectively. Solar prominences are a spectacular example of MHD systems in a quasi-
stable equilibrium. Prominences are enormous tubes of cool dense material, held up high in the corona,
presumably by twisted magnetic fields. We say “quasi-stable” because they are sometimes observed
to last for weeks, or months. Sometimes prominences are observed to “erupt”, i.e., to appear to burst
upwards, disappearing rapidly, in the space of a few hours. Such eruptive prominences are obviously
due to some large scale (MHD?) instability. The reason for the transition from stability to instability is
not completely clear, but is probably due to a rearrangement of the magnetic fields at the base of the
corona.
We can set ∂/∂t = 0 in the MHD equations, and then examine the resulting solutions which are
called magnetohydrostatic (MHS) equilibria. Since much of these studies are concerned with coronal
structures we include the gravitational term:
1
∇p − ρg = (∇ × B) × B (4.61)
µ0
kb
p = ρT (4.62)
m
∇·B = 0 (4.63)
Along the magnetic field there is no contribution from the magnetic force, and so we have a hydro-
static balance between pressure gradients and gravity. Comparing the gas pressure p to the magnetic
pressure B2 /2µ0 defines the plasma beta
p 2 c2s
Plasma Beta β= = (4.64)
B2 /2µ0 γ vA2
Figure 4.8: Friedrichs Diagram showing MHD wave mode phase speeds as a function of angle between
k and B0 . The distance r to a point P in the direction of k is proportional to the phase speed of the waves
propagating in that direction. The case on the left corresponds to c2s vA2 for which the fast mode is
essentially a modified sound wave (in the limit vA2 → 0 the Alfvén and slow modes shrink to the origin
and the fast mode “ripple” becomes circular). The case on the right corresponds to vA2 c2s .
For β 1 the magnetic field behaviour is dominated by the gas pressure, and the Lorentz force can
be neglected. For β 1 the magnetic field dominates, and the pressure may be neglected (this is case
for solar corona, usually). When β 1 and L the typical scale length of any structures is such that
L < h where h is the pressure scale height (the distance over which the pressure falls by a factor e due to
gravity), then the gravitational force can be neglected. In this case (typical of active regions on the Sun)
one finds so called force-free equilibrium where
Force Free Equilibrium (∇ × B) × B = 0 (4.65)
A more restrictive case is when one further specifies that there is zero current. Such current-free
equilibria obey
∇×B = 0 (4.66)
A second solution exists to (4.65) in which the electric current j is parallel to the magnetic field B. The
force-free equilibrium equation in this case is deceptively simple, especially when it is noted that it is
equivalent to
∇ × B = αB (4.67)
where α = α(r). In fact a major area of research is finding suitable solutions for α. Of course when
α = constant, the force-free equilibrium equation becomes much easier to solve.
MHS Example: Cylindrically Symmetric Flux Tube

The subject of MHD equilibria is too vast to go into too much detail here. Various texts are devoted to
the subject, and textbooks which deal with the solar atmosphere are useful starting points.
For now, let us just explore briefly the nature of equilibria by considering a cylindrically sym-
metric flux tube. We work in (r, φ, z) coordinates and assume the magnetic field to be of the form
B = (0, Bφ (r), Bz (r)) while the gas pressure p = p(r) is similarly only a function of the radial distance.
Our choice of B satisfies ∇ · B = 0 without further ado (see Equation A.19), so the MHS equations,
ignoring gravity, reduce to just " #
d B2φ + B2z B2φ
p+ + =0 (4.68)
dr 2µ0 µ0 r
The first term behaves like a pressure gradient, and the second like a tension. It is clear that an isolated
flux tube, in which one expects the total pressure to decrease from the interior, requires an azimuthal field
Bφ to have any chance of an equilibrium. Thus magnetic flux tubes tend to be “twisted”. Equation 4.68
is only one equation for three unknowns, therefore we have the freedom to specify two of them, and
solve for the third.
4.6.3 MHD Instabilities

Instabilities arise when an equilibrium represents a system possess free energy in the sense that, for
some small perturbation, the perturbation grows and the system departs further from the equilibrium
state toward one of lower energy.
There are many different kinds of MHD instabilities. They include the following: interchange
modes, in which field lines are wrapped around plasma in a concave manner; Rayleigh-Taylor instability,
in which plasma is supported by a field against gravity, which may create structure in prominences;
sausage and kink modes of a flux tube; Kelvin-Helmholtz instability, in which plasma flows over a
magnetic surface; resistive modes of a sheared magnetic field, which drive reconnection (see Chapter 6);
convective instability when a temperature gradient is too large, which can concentrate flux tubes in
the photosphere; radiative instability, which creates cool loops and prominences up in the corona; and
magnetic buoyancy instability of a magnetic field in a stratified medium in which the density decreases
with height. Buoyancy causes flux tubes to rise through the solar convection zone.
In each case, the question of nonlinear development and saturation of the instability is important
(and difficult). In general, instability calculations proceed by linearising the governing equations, as we
did in Section 4.6.1 except now the background state is not uniform. The resulting dispersion relation
gives rise to complex frequencies ω = ωr + iγ which results in growing solutions for γ > 0. [In the MHD
equations, in fact, it is possible to show that ωr ≡ 0.]
4.7 Dynamos
Magnetic fields abound in most, if not all, astrophysical environments. So the obvious question is
“Why?”. Two answers spring to mind: Either the universe was born with “primordial” magnetic fields,
which have been decaying ever since, or else something is generating them. In fact, it is possible that
primordial fields are still lurking around, but there is ample evidence that fields can be generated. In the
4.7. DYNAMOS 51
laboratory, simple circuits and currents give rise to magnetic fields due to Ampere’s Law (Equation 2.2).
In plasmas, however, fields are more often induced via electromotive effects embodied in the induction
equation (4.13). This forms the basis for investigations into the generation of magnetic fields, generally
known as “the dynamo problem.”
∂B
= ∇ × (V × B) + η∇2 B (4.69)
∂t
where we have introduced the magnetic diffusivity η ≡ 1/(µ0 σ) to emphasize the role of this term. As
we saw in Section 4.2 the last term in this equation represents the magnetic diffusion while the middle
term the convection of the field by the plasma bulk flow. Then the decay of the magnetic field due to
diffusion (which also leads to dissipation of the field) gives rise to a characteristic decay time τ found
from (4.69) as
B B
∼η 2 (4.70)
τ L
Using standard (collisional) values to evaluate τ gives long, but not infinite, timescales. In the Earth’s
core, τ ∼ 3 × 105 years; in the Sun, τ ∼ 1011 years, only marginally longer than the age of the Sun. This
simple calculation suggests that the Earth needs something to replenish its magnetic field, but perhaps
the Sun doesn’t. However, we know that the Sun’s field reverses on a short timescale (11 years), as
does the Earth’s, though not nearly so frequent nor regular. Thus some active process is generating
these fields. Dynamos really do exist. The only problem is explaining them from our mathematical and
plasma physical understanding.
4.7.1 The Dynamo Problem

The full dynamo problem consists of using the MHD equations to show how fluid motions V and mag-
netic fields B can be self-consistent and self-maintaining. The motion induces magnetic fields via the
induction equation given above, while the fields in turn influence the motion through the MHD momen-
tum equation (4.2). This problem is highly nonlinear and difficult to treat, even numerically. Even the
induction equation (4.69) on its own is complicated by the nonlinear convective term.
A compromise to this full problem is to consider the kinematic dynamo problem in which the flow
is prescribed and the consequences for the magnetic field explored. This makes (4.69) linear in the
unknown field behaviour and enables some progress to be made which, hopefully, sheds light on the full
dynamo problem.
Known astrophysical fields tend to be dipolar in structure and aligned roughly with a symme-
try/rotational axis, so it would be natural to seek solutions with these properties. Unfortunately, a result
due to Cowling, known as Cowling’s Theorem, intervenes. Cowling showed that no steady, axisymmet-
ric solution for a dynamo with laminar fields and flows is possible! Since Cowling’s theorem applies to
the exact field, it is possible that there are solutions for which some average field and flow are axisym-
metric. This has spawned a considerable body of literature on the subject of mean field electrodynamics,
in which the properties of fluctuations due to turbulence are inserted and investigated. We shall return to
this subject in Section 4.7.3.
4.7.2 Qualitative Dynamo Behaviour

Although the mathematics of dynamo behaviour quickly becomes complicated, the qualitative process
can be illustrated rather simply and consists of the following sequence of operations, sketched in Fig-
ure 4.9:
Figure 4.9: Steps in the idealised steady dynamo process
Stretch the field lines to lengthen them. This process does work on the field, transfering flow energy to
the magnetic energy density.
Twist the stretched field, to change the direction of the stretched field toward that of the original field.
Further twists or folds then amplify the original field. This is the key step.
Diffuse the field across the small scales which are imposed in the above steps to return the system
toward its starting point.
4.7.3 Mean Field Kinematic Dynamos

We restrict ourselves here to perhaps the simplest mathematical formulation of the dynamo problem.
This is a pure kinematic approach, in which the velocity field is prescribed and we seek the resulting
magnetic field by looking at the induction equation. We proceed by decomposing all quantities into their
mean values (subscripted “0”) and their fluctuations (subscripted “1”). Thus we write
B = B0 + B1 (4.71)
V = V0 + V1 (4.72)
where B0 ≡< B >, the average being over space, time, or ensemble (we shall not need to distinguish
amongst these) so that < B1 >= 0. This is reminiscent of the linearisation we performed in Section 4.6.1
except that here there is no assumption that the fluctuating quantities B1 , V1 are small. Substitution into
the induction equation (4.69) leads to
∂
(B0 + B1 ) = ∇ × [(V0 + V1 ) × (B0 + B1 )] + η∇2 (B0 + B1 ) (4.73)
∂t
Taking the average of this equation leads to the following equation which governs the mean magnetic
field B0 :
∂B0
= ∇ × (V0 × B0 ) + ∇ × (< V1 × B1 >) + η∇2 B0 (4.74)
∂t
4.7. DYNAMOS 53
Thus we can see that the fluctuations in V and B give rise to an extra electromotive force E ≡ < V1 × B1 >
which can contribute to the inducted magnetic field. The details of E depend on the nature of the fluctua-
tions, usually assumed to be turbulent. In particular, if the mean fields are axisymmetric, the fluctuations
cannot be in order to avoid the constraints of Cowling’s Theorem. An exposition of the momentum
equation (to determine V1 ) and difference between (4.73) and (4.74) (to determine B1 ) is beyond the
scope of this text. However, rather general arguments lead to the conclusion that E is of the form
E = α · B0 + β : ∇B0 + . . . (4.75)
Taking the simplest case, in which the tensor α is isotropic and β is a scalar constant leads to
∂B0
= ∇ × (V0 × B0 ) + ∇ × (αB0 ) + (η + ηT ) ∇2 B0 (4.76)
∂t
where we have introducted the turbulent diffusivity ηT which results from the β term in E . Given the
typically long decay times based on classical diffusivity, this role for the turbulence in the medium is
essential if a balance between field generation and dissipation is to be reached. Equation 4.76 forms
the starting point for many investigations of dynamo action. In terms of the dynamo steps depicted in
Figure 4.9, the first term on the right hand side embodies the stretching of the field by the mean flow,
while the second term contains the “α-effect” twisting due to the helicity in the turbulence.
4.7.4 α − ω Solar Dynamo

The solar dynamo represents a key challenge for any dynamo theory. The simplest mathematical ap-
proach begins with (4.76) and makes the following further simplifications/assumptions:
• The mean fields B0 and V0 are axisymmetric, i.e., they have no φ dependence in a spherical polar
representation. In particular, we take B0 = B0 (r, θ)
• The mean velocity field is a radially dependent, differentially rotating fluid, i.e., V0 = ω(r, θ)r sin θφ̂
The first term on the right hand side of (4.76) can be expanded by using (A.7) into four terms. Note
firstly that ∇ · B0 = 0 thanks to maxwell and that ∇ · V0 = 0 using (A.23) with the knowledge that V0 has
only a φ̂ component and is independent of φ. The remaining two terms combine using (A.26) to give
∂ω ∂ω
" #
∇ × (V0 × B0 ) = φ̂ sin θ rBr + Bθ (4.77)
∂r ∂θ
This result may also be obtained more directly by first evaluating A ≡ V0 × B0 and then using (A.24)
followed by some re-arrangement making use of ∇ · B0 = 0 as evaluated in spherical polars.
The result of all this is
∂B0 ∂ω ∂ω
" #
= φ̂ sin θ rBr + Bθ + ∇ × (αB0 ) + (η + ηT ) ∇2 B0 (4.78)
∂t ∂r ∂θ
The first term on the right hand side embodies the ω-effect, which winds up an initial poloidal field
to generate a toroidal (φ̂) one. The second term describes the α-effect, in which the turbulent motions
result in a twisting of the field. In the solar case, a net twist (as opposed to equal and cancelling twists) is
accomplished by the coriolis effect on rising flux in the stratified solar atmosphere that is not symmetric
with that on falling flux tubes. These effects are illustrated in Figure 4.10.
Figure 4.10: Illustration of the ω-effect (left) which stretches poloidal field lines and generates toroidal
ones due to the differential rotation of the Sun, in which the equatorial regions rotate faster than the
polar regions and the α-effect (right), in which rising flux tubes in the turbulent medium are twisted by
coriolis effects.
4.7.5 Astrophysical Dynamos
Solar Dynamo
We have already explored many aspects of the solar dynamo. The 11-year reversals pose a major chal-
lenge to dynamo theory. Additionally, within that cycle, concentrations of strong magnetic field, such
as sunspots, appear at mid-latitudes during the early phases of the cycle, but the zones of activity appear
increasingly at lower latitudes as the cycle progresses. This gives rise to a “butterfly” pattern shown in
Figure 4.11.
Another feature of solar magnetic fields is the polarity of emerging flux regions. This polarity
follows a regular pattern, with the leading polarity opposite in the opposite hemispheres. The sense
reverses with each solar cycle reversal, as shown in Figure 4.12. These facts, known as Hale’s Law,
are qualitatively consistent with the toroidal field generated due to differential rotation sketched in left
portion of Figure 4.10, which then rises due to magnetic buoyancy.
Disk Dynamos
Astrophysical systems involving disks, such as accretion phenomena, are also sites of dynamo action. In
these cases, kinematic instabilities give rise to turbulent fluid motions which, in turn, generate fields via
dynamo action. These fields, however, are purely turbulent, with no large-scale mean field. Nonetheless,
the turbulent diffusivity (for both matter and fields) plays a critical role in the evolution of these systems
in terms of angular momentum transport and accretion rates.
4.8. EXERCISES 55
Figure 4.11: Appearance of strong flux regions at different solar latitudes (top) with time, which is
related to solar activity as measured, e.g., by total sunspot numbers (bottom). The equatorward drift of
emerging flux gives rise to a characteristic “butterfly” diagram.
4.8 Exercises
1. A perfectly conducting plasma has a velocity −V0 x̂ and is threaded by a magnetic field B0 ŷ. An
obstacle stands in the flow, and causes a deceleration of the plasma within the region R such that
!
dV y V0
= + 1 − 0.9

dx L L
Assume that the flow field is fixed once the plasma moves into the region x < 0, i.e., that V (x <
0, y) = V (x = 0, y), and that the flow remains in the negative x-direction (i.e., there is no deflection
of the flow by the obstacle). Take the ẑ direction to be invariant. (see Figure 4.13)
(a) By applying your knowledge of the frozen-in flux principle, sketch the magnetic field struc-
ture in the x − y plane.
(b) What is the magnetic field strength at (0, 0)?
(c) What has happened to the kinetic energy of the flow passing through the region R? (How
does conservation of energy apply?)
(d) What would really happen in the region x < 0 and why?
2. A uniform perfectly conducting spherical plasma cloud of mass density ρ0 , total mass M and
radius R0 is rotating at an angular frequency Ω0 . The cloud is permeated by a magnetic field
Figure 4.12: Magnetic polarity at various solar latitudes as a function of longitude during two solar
cycles, illustrating Hale’s Law. Note that within each cycle the leading polarity is opposite in opposite
hemispheres, and that this polarity reverses from one cycle to the next.
which is everywhere aligned with the rotation axis, has a magnitude B0 at the equatorial surface,
and increases linearly to twice that value at the rotation axis.
(a) This cloud now suffers a collapse to a radius R1 < R0 . The cloud remains uniform so that
its total (conserved) angular momentum throughout the collapse phase is given by 52 MR20 Ω0 .
Find the resulting mass density, ρ1 , and angular velocity, Ω1 . Hence deduce that, at the
equatorial surface the centrifugal force per unit volume, ρ1 Ω21 R1 , has increased by a factor
(R0 /R1 )6 over its initial value.
(b) Use simple flux freezing arguments (without performing any explicit integrations) to esti-
mate the factor by which the magnetic field, B1 , at the equatorial surface has increased.
Hence, assuming that the magnetic profile remained roughly linear, estimate, e.g., to within
a factor of order unity, the magnetic pressure force per unit volume at the equatorial surface
and give an expression for this force in terms of its initial value.
(c) Finally, by comparing the centrifugal and magnetic forces found above, show that the mag-
netic forces increase relatively less, by a factor R1 /R0 , than the rotational ones, so that the
4.8. EXERCISES 57
Figure 4.13: Flow hitting an obstacle.
conservation of angular momentum represents a more serious inhibitor of further collapse

than the build up of magnetic pressure.
3. Consider a uniform plasma of density ρ threaded by a uniform magnetic field B. By making an

analogy to the case of small perturbations on a string in which the mass per unit length is ρ and
the tension T is equivalent to the magnetic field tension B2 /µ0 , show that the transverse waves on
the string propagate with a phase velocity equivalent to the Alfvén speed.
4. Fill in the steps to derive the general MHD dispersion relations (4.59) and (4.60) from (4.49).
[Hint: For (4.60) dot (4.49) with k and B0 and then work with the new variables (k · δV) and
(B0 · δV).]
5. Evaluate the dispersion relation for fast and slow magnetosonic waves (4.60) in the regime c2s vA2
√ 1 1
by expanding the radical [ 1 − x ≈ 1 − x − x2 . . . for |x| 1]. Show that
2 8
(a) the slow mode is guided in the sense that the group velocity ∂ω/∂k is directed along the
background magnetic field.
(b) that the pressure perturbations δp are out of phase with the perturbations in the magnetic
field pressure δ|B2 |/2µ0 ≡ δB · B0 /µ0 . This is a bit challenging. Try dotting (4.49) with k
and (4.46) with B0 . Then identify the terms in the first dot product which are the same form
as the second. The remaining terms are all proportional to k · δV and so you can derive a
relationship between δp and δB · B0 . Use your result for the dispersion relation to show that
the proportionality constant is negative. [You can now easily also show that in the case of
the fast magnetosonic wave the magnetic and thermal pressures are in phase.]
6. Consider the equilibrium of an isolated cylindrically symmetric flux tube of radius r0 immersed

r
in a uniform field Bz ẑ. Given that the plasma thermal pressure varies as p(r) = p0 1 − show
r0
that the azimuthal magnetic field must vary according to
B2φ r
= p0 (4.79)
2µ0 3r0
[You will need to solve Equation 4.68 or at least show that this is a valid solution in the region
r ≤ r0 . You will also need to assign appropriate boundary conditions at, say, r = 0.] Also, find the
electric current density j = ∇ × B/µ0 for this configuration.
7. More questions to follow (dynamos)

Chapter 5
The Solar Wind
5.1 Introduction
The solar wind is a flow of ionized solar plasma and an associated remnant of the solar magnetic field
that pervades interplanetary space. It is a result of the huge difference in gas pressure between the solar
corona and interstellar space. This pressure difference drives the plasma outward, despite the restraining
influence of solar gravity. The existence of a solar wind was surmised in the 1950’s on the basis of
evidence that small variations in the Earth’s magnetic field (geomagnetic activity) were produced by
observable phenomena on the sun (solar activity), as well as from theoretical models for the equilibrium
state of the solar corona. It was first observed directly and definitively by space probes in the mid-l960’s.
Measurements taken by spacecraft-borne instruments since that time have yielded a detailed de-
scription of the solar wind across an area from inside the orbit of Mercury to well beyond the orbit of
Neptune. Our interest in this distant and tenuous plasma stems from two important aspects of solar wind
research.
The first of these concerns the role of the solar wind in the interdisciplinary subject known as solar-
terrestrial relations. The solar wind is significantly influenced by solar activity (or, in physical terms, by
changes in the solar magnetic field) and transmits that influence to planets, comets, dust particles, and
cosmic rays that “stand” in the wind. The origin of the solar influence through interaction of the solar
magnetic field with the expanding coronal plasma is a major topic in present-day solar-wind research.
The second important aspect of solar-wind research concerns the physical processes that occur in its
formation and expansion from the hot solar corona to the cool and far more tenuous regions of the outer
solar system. This expansion takes the magnetized plasma through huge variations in its properties; for
example, collisions among ions or electrons in the expanding plasma are frequent in the corona, but
extremely rare in interplanetary space. Thus the physics of this stellar plasma system can be examined
under a wide variety of conditions, some of which are difficult or impossible to reproduce in terrestrial
laboratories or in the immediate vicinity of the Earth. However, the solar wind is accessible to space
probes, and its properties can be measured and its physical processes studied at a level of detail which
is impossible for most astrophysical plasmas.
5.2 Description
Most of our observations of the solar wind have been made by spacecraft near the orbit of the Earth.
Typical values for solar wind parameters at this distance (i.e., 1 AU) are given in Table 5.1. The solar
wind exhibits considerable variations, with, for example, the flow speed often less than 300km/s or
greater than 700km/s.
The embedded magnetic field (the interplanetary magnetic field or IMF) in the Earth’s vicinity lies,
59
60 CHAPTER 5. THE SOLAR WIND
Table 5.1: Typical Solar Wind Parameters at 1AU

Parameter Value at 1AU
Proton density 6.6 cm−3
Electron density 7.1 cm−3
He2+ density 0.25 cm−3
Flow speed ( radial) 450 km s−1
Proton temperature 1.2 × 105 K
Electron temperature 1.4 × 105 K
Magnetic field 7 nT
on average, in the ecliptic plane, but at approximately 45◦ to the Earth-Sun line. We will consider simple
models to account for these observed properties of the solar wind and the IMF.
5.3 Why is there a solar wind? - A simple model

The supersonic flow of the solar wind comes about from the conversion of thermal energy in the corona
(at low velocities) to kinetic energy (high velocities) of radial outflow. We will investigate this conversion
using a very simple model of the solar wind. This model makes a large number of assumptions, but
nevertheless has most of the important features of more complicated models.
We use the MHD equations in the same notation as in Chapter 4 and look for steady state solutions
∂
with = 0. The equations for conservation of mass and momentum become
∂t
∇ · (ρV) = 0 (5.1)
ρ (V · ∇) V = −∇p + j × B + ρFg (5.2)
where Fg is the gravitational force per unit mass
GM
Fg = − r̂ (5.3)
r2
We now make three major assumptions
1. We assume radial symmetry, so that any flow is strictly radial, and all quantities depend only on
the radial distance r.
2. We also neglect the magnetic force term in the momentum equation, so that we only examine the
effect of the pressure gradient that drives the flow, and not any back-reaction from the magnetic
field. (This might seem strange, given that the importance of magnetic fields has been stressed,
but it makes the problem much more attractive!)
3. We treat the plasma as isothermal (i.e., γ = 1, constant temperature), with the pressure given by
the ideal gas law p = nkb (Te + Ti ) = 2nkb T and where the mass density is related to the particle
number density by ρ = nm (m = me + mi ). The assumption of constant temperature is equivalent
to assuming infinite thermal conductivity.
The assumption of radial symmetry has the following consequences
V = V (r)r̂ (5.4)
5.3. WHY IS THERE A SOLAR WIND? - A SIMPLE MODEL 61
dp
∇p = r̂ (5.5)
dr
1d
∇ · (ρV) = ρV r2

2
(5.6)
r dr
dV
ρ (V · ∇) V = ρV r̂ (5.7)
dr
So, our simplified model reduces to
d
ρV r2

= 0 (5.8)
dr
dV dp ρGM
ρV = − − (5.9)
dr dr r2
ρ
p = 2kb T (5.10)
m
5.3.1 Static Atmosphere: V (r) = 0

Since we are trying to investigate the formation of the solar wind, it may seem perverse to start with the
case of a static atmosphere. However, as well as being the historical starting point, it is revealing to find
out why a static solar atmosphere is not possible.
Setting V = 0 automatically satisfies Equation 5.8 for mass conservation, and the momentum equa-
tion (5.9) becomes
d p ρGM
0=− − (5.11)
dr r2
Using (5.10) with T =constant we can find an equation for p only
1 dp GM m 1
=− (5.12)
p dr 2kb T r2
which has the following solution

GM m 1
ln p = +K (5.13)
2kb T r
where K is constant. Suppose that the pressure at the base of the corona, r = R, is p = p0 . Then we find
the full expression for the pressure as a function of radial position
" !#
GM m 1 1
p(r) = p0 exp − (5.14)
2kb T r R
For r > R, p < p0 , i.e., the pressure decreases with radial position.
In the case of a “shallow” atmosphere such as the Earth’s, where r − R R , then the pressure
decrease becomes exponential of the form p = p0 e−h/λ . But the problem we encounter for the Sun’s
corona is that the atmosphere is not shallow, so for large r
" #
GM m 1
p → p0 exp − (5.15)
2kb T R
Using typical values for the solar corona, with T = 106 K, one finds that the exponential term is only
a factor 3 × 10−4 below the coronal pressure, and this is much higher than the observed “interstellar”
pressure. One is therefore led to the conclusion that a static solar atmosphere cannot be in equilibrium
with the interstellar medium, and so we must look for solutions with an outflow velocity.
5.3.2 Solar Wind Solutions: V (r) 6= 0

The history of the prediction and eventual confirmation of the existence of the solar wind is an interesting
one. In the late 1950’s when Parker proposed a continual, supersonic flow from the corona, the result
of the static atmosphere was known, while the possible incompatibility with the interstellar medium
pressure was not really known for certain. However, it was acknowledged that there had to be a way
for the sun to influence the Earth in a fairly rapid and powerful way, since large geomagnetic storms
were often seen associated with, but a few days after, a solar flare. It was thought that the flare ejected
a stream of protons in some collimated beam (the solar “corpuscular radiation theory”), but of course
anybody with knowledge of plasmas would know that this single species beam would be very unstable.
On the other hand Biermann had been making observations of the plasma tails of comets, and since the
early 1950’s had been emphasizing how these were permanent features, and that they always pointed
away from the sun. Parker was aware of these observations, and decided that by proposing a steady,
flowing solar wind he could explain both the orientation of the cometary plasma tails and also the way
in which perturbations at the sun (e.g., flares) could be carried by the flow to the Earth, where they could
perturb the magnetic field as measured at the surface of the Earth.
We now follow, in simplified form, the argument that Parker suggested.
Integrating the mass conservation equation (5.8) finds ρV r2 = C where C is a constant, or equiva-
lently
I = 4πr2 ρV (5.16)
where I is the mass flux through a sun centred sphere of radius r, which is constant when the flow is
independent of time.
Again assuming an isothermal expansion (5.10) for p the momentum equation (5.9) becomes
dV dn ρGM
ρV = −2kb T − (5.17)
dr dr r2
Dividing by ρ ≡ nm gives
dV 2kb T 1 dn GM
V =− − (5.18)
dr m n dr r2
Our aim is to reduce this to an equation for V (r), so from the mass conservation equation
I 1
n= (5.19)
4πm V r2
Differentiating !
dn I 1 dV 2
= − − (5.20)
dr 4πm r2V 2 dr V r3
Simplifying, and substituting back into the momentum equation:
dV 4kb T 2kb T 1 dV GM
V = + − (5.21)
dr mr m V dr r2
And rearranging gives

2kb T 1 dV 4kb T GM
Parker’s Solar Wind Equation V2 − = − 2 (5.22)
m V dr mr r
This ends our search for a governing equation for the radial outflow. It is not too difficult to solve this
ordinary differential equation. But it is interesting to look for the behaviour of solutions to this equation
in a qualitative fashion, by asking how left hand side (LHS) and right hand side (RHS) can balance.
Considering the RHS: for the observed coronal temperatures one finds that the RHS is negative, i.e.
GM 4kb T
2
> (5.23)
r mr
That is, the corona is gravitationally bound. For increasing r, gravitational term decreases faster than the
thermal term. So the RHS for small r is negative, but becomes positive after passing through zero at the
critical radius rc at
GM m
Critical Radius rc = (5.24)
4kb T
Now consider the LHS. Observations show that at small r the flow velocities V (r) are smaller than the
thermal velocity, so
2kb T
V2 − <0 (5.25)
m
But the RHS at this time is also negative, so in order to balance one must have dV /dr > 0, which
implies that V increases with r. So at small r the flow is small but increases with radius. But as r
increases beyond rc there are two possible solutions which depend on what happens at rc . At rc the RHS
is zero, so either dV /dr = 0 (and V has a maximum or minimum) or V 2 − 2kb T /m = 0 (and dV /dr can
remain positive). These possibilities are sketched in Figure 5.1
The solution which has dV /dr passing through zero at r = rc must have dV /dr < 0 for r > rc , so
that the flow becomes small at large radial distances. The problem with this case is that it resembles the
static atmosphere, which does not balance with the known interstellar pressure. Thus we are led to reject
this type of solution.
The solution with V 2 − 2kT /m = 0 at r = rc , on the other hand, maintains dV /dr > 0 and has a flow
speed which continues to increase with radial distance. If V (r) increases with r, then the number density
I 1
n(r) = 2
(5.26)
4πm r V (r)
decreases with r, and so the pressure p = 2nkb T becomes increasingly small at large radial distances, so
that there is indeed the possibility of coming into balance with the interstellar pressure.
p critical radius is especially important in the following respect. The flow velocity at r = rc is
The
V = 2kb T /m. Recall that the speed of sound is given by
γp 2kb T
c2s = =γ (5.27)
ρ m
Figure 5.1: Possible solutions to Parker’s solar wind equation (5.22). The breeze solution is close to
the static atmosphere case and asymptotes to a finite pressure at large r. Only the solar wind solution,
which passes through the sonic point at the critical radius rc matches the solar and interstellar bound-
ary conditions. Interestingly, Parker’s equation only involves the square velocity, and so also includes
accretion-type solutions, such as the solution which is sub-sonic at large r and increases through the
sonic point to large velocities close to the star.
Figure 5.2: Isothermal radial expansion speed solutions V (r) for different coronal temperatures.
Figure 5.3: Loci of a succession of fluid parcels emitted at constant speed from a source rotating with
the sun.
For an isothermal medium, γ = 1, and one can identify the critical radius at which the solar wind solution
becomes supersonic. Observations do indeed reveal that the solar wind is supersonic at 1AU. Figure 5.2
shows solar wind solutions for various temperatures.
Much research effort has been spent making more sophisticated models of the solar wind, by in-
cluding the effects of the magnetic field, thermal conductivity, the presence of alpha particles (ionized
helium, which is an important minority species), and so on. Unfortunately, the increasing sophistication
of the models has not resulted in any one model that can predict every feature of the solar wind at 1
AU (at the Earth’s orbit). It seems necessary that additional heating must take place in the solar wind in
the corona, so that the simple model above is not completely correct. But the nature of this additional
heating process, although probably originally from the photospheric motions, has not yet been identified.
5.3.3 The Interplanetary Magnetic Field (IMF)

Since the Magnetic Reynold’s (4.14) number in the solar wind is high, we expect that the Flux Freezing
Theorem will be valid, and the magnetic field at the surface of the sun will be carried out into interplan-
etary space by the solar wind.
The situation is complicated by the fact that the sun (and hence the foot points of the magnetic field
lines) rotates every 25.4 days. The apparent period of rotation at the Earth is about 27 days. This is
an angular frequency of Ω ≈ 2.7 × 10−6 rad/s. But once a plasma “parcel” has left the sun, it moves
approximately radially, due to its high radial speed. However, the magnetic field line which passes
through it is tied to the original coronal position which is rotating. The effect is to produce a spiral
pattern, like the streams of water from a rotating water sprinkler.
Consider separate plasma parcels emitted from the same location at the corona. They move out
radially, but they are joined by the same field line; plasma elements initially on the same field line,
remain on that field line for all time (flux freezing). This is illustrated in Figure 5.3. Thus the IMF
has spiral structure, which tends to wind tighter as one goes further from the sun (i.e., perpendicular to
the axis of solar rotation), and which is less tightly wound as one passes from the ecliptic plane toward
higher heliographic latitudes. The last point can be seen by considering plasma parcels emitted from a
pole of the sun, in which case the footpoint stays aligned with the plasma parcel at all times.
What is the equation describing this spiral structure? We choose a corotating frame, angular fre-
quency Ω, so that the sun is stationary. The solar wind radial velocity Vr = V (r) is the same as before,
but in addition there is an azimuthal velocity Vφ = −ωr. (We are here just considering what happens in
the ecliptic plane.)
In order for field lines to connect the same plasma parcels we must have
Bφ Vφ Ωr
= =− (5.28)
Br Vr Vr
which, since Vr = dr/dt and Vφ = rdφ/dt, leads to
dφ Ω
=− (5.29)
dr Vr
If Vr is constant (which is a pretty good approximation outside the critical radius rc as can be seen, e.g.,
in Figure 5.2) we can solve this ordinary differential equation to find the equation of the field line
Vr
r = R− (φ − φ0 ) (5.30)
Ω
where φ = φ0 at r = R. This spiral is known as an Archimedean spiral although, in the context of the
solar wind, it is known as the Parker spiral.
What are the components of the IMF? We need to satisfy ∇·B = 0, which in spherical polar geometry
(and assuming spherical symmetry for the moment)
1 ∂
∇·B = r2 Br

(5.31)
r ∂r
2
so that Br ∝ r−2 . Also the assumption of constant solar wind speed V is fairly good throughout the
heliosphere, so that we can choose a reference point r0 , and writing
!2
r0
Br = Br (r0 ) (5.32)
r
then, from (5.28)

Ω r02
Bφ = − Br (r0 ) (5.33)
Vr r
So if the angle between the magnetic field and the sun-Earth line is ψ, then we have
Bφ Ωr
tan ψ = − = (5.34)
Br Vr
At the Earth’s orbit ψ is about 45◦ , and gets larger further away from the Sun.
Transforming back into the inertial (nonrotating) frame one can see that there is an induced electric
field, because of the angle between the flow velocity and magnetic field
E = −V × B = +V Bφ θ̂ (5.35)
In the above we have assumed that the flow is radial. The azimuthal velocity imparted directly by the
solar rotation is negligible in the low coronal, but perhaps the magnetic field can influence this behaviour
since the flow and field are frozen together. In a strong field, magnetic forces due to the field threading
from the corona out to larger distances would force the wind to co-rotate with the Sun. This proceeds, if
5.4. THE REAL SOLAR WIND 67
at all, only until the flow energy density becomes dominant over that of the field (beyond which the field
goes where the flow takes it!) So the field winds into the Parker sprial in regions where
B2 1
ρV 2 (5.36)
2µ0 2
√
that is, where V vA , where vA = B/ µ0 ρ is the Alfvén speed. We can define the Alfvén radius RA
where V = vA , which for typical values is at about 50R , or 0.25AU. Inside the Alfvén radius the field
is strong enough to control the flow, and cause the solar wind to corotate with the solar wind. Outside
the Alfvén radius the field wraps up according to the Parker spiral.
While such co-rotation is not important for the present dynamics of the solar wind, this calculation
does show that, for example, the angular momentum shed by stellar winds can be much larger than what
might be expected due to the surface rotation, since escaping wind particles are rigidly rotating out to
the Alfvén radius and thus carry away (RA /R? )2 more angular momentum.
5.4 The Real Solar Wind

So far we have considered only the most basic of solar wind models: one-fluid, spherically symmetric,
and steady. The real solar wind is a much richer medium, providing numerous examples of important
astrophysical phenomena on all scales.
5.4.1 Structure
Large Scale Structure
On the largest scales there are a variety of important structures found in the solar wind. These include:
Sectors The various regions of magnetic polarity at the Sun give rise to alternating regions of “toward”
and “away” fields since the heliospheric current sheet which divides these is inclined relative the
the ecliptic and highly warped. During periods of solar minimum the interplanetary field seen at
the Earth is dominated by two recurring sectors (one toward, one away) per solar rotation. During
periods of maximum activity, the IMF is much more complex, with four or more recognisable
sectors and many more transient phenomena.
Fast/slow Streams Related to the magnetic structure are solar wind streams whose properties are de-
pendent on the coronal conditions from which the originate. Open, coronal hole regions which
predominate during solar minima give rise to high speed (600-800 km/s or more) streams. Solar
wind emerging around closed loop regions (“streamers”) is slower and more highly variable.
Co-rotating Interaction Regions CIRs are formed near and beyond 1AU due to the interaction of slow
and fast solar wind streams. Due to the solar rotation, a slow solar wind parcel travelling radially
outward will eventually get caught up by a fast parcel emitted from a different region on the
Sun which has rotated into the same radial direction. This presents a classic stream-on-stream
problem and results in a pair of forward/reverse shocks (see Chapter 7). In the outer heliosphere,
these interactions merge and process the intervening solar wind.
Coronal Mass Ejections The Sun, especially during periods of high solar activity, releases large clouds
of solar material which pass through the corona and accelerate to high speeds. These coronal mass
ejections (CMEs) can fill an entire spherical quadrant with highly structured magnetic field and
dense plasma. CMEs are the main drivers of geomagnetic activity.
Small Scale Structure

On plasma kinetic scales, the solar wind is a rich, collisionless medium with multiple plasma compo-
nents (protons, electrons, alpha particles, . . .) and magnetic field fluctuations. The study of these aspects
sheds considerable light on the coronal heating and solar wind acceleration problems (which despite
decades of research are far from understood) and also build expertise which can be applied to other
astrophysical objects where winds and particle acceleration are believed to exist. Examples of these
phenomena include
Waves The solar wind is a highly turbulent medium with a mixture of low frequency, MHD-like waves
as well as high-frequency waves driven by accelerated particle beams and non-Maxwellian particle
distributions.
Discontinuities The structure of discontinuities separating different plasma regimes (e.g., the fast and
slow streams discussed above) is vital to understand how such regimes interact and under what
circumstances the “cell model” (see Section 4.4) is appropriate.
Shocks Shocks abound in astrophysical environments, and the heliosphere offers an unrivalled labo-
ratory for the study of collisionless shock formation, particle acceleration, and accompanying
turbulence. Much of the material in Chapter 7 is based on observations at the Earth’s bow shock
and interplanetary shocks. Shocks are formed at stream interfaces and where the solar wind en-
counters obstacles with extended atmospheres/magnetic fields, such as planetary bow shocks and
comets.
5.4.2 Other Physics

In our treatment of the solar wind, we have idealised the solar wind problem in order to expose its essen-
tial ingredients. Many important ingredients and problems have been neglected. Most were anticipated
by Parker in his original papers written in the late 1950’s. Research, both analytical and increasingly
numerical, has sought to address many of these problems, with mixed success. Here is an incomplete
list of topics
Coronal heating - beyond the obvious question of why the million degree corona exists at all above
the few thousand degree solar surface given the solar energy source deep in the interior, it seems
some extended heating is required.
Trans-collisional physics - the lower corona is collision-dominated; the outer corona/solar wind nearly
collisionless. Treating the intervening regions is extremely difficult.
Thermal conduction - the solar wind is essentially the conversion of thermal energy to bulk flow, so
the thermal transport processes are vital. Conduction in turbulent, collisionless media is not at all
understood.
Multi-species - many species of ions are present in small amounts in the solar wind. Curiously, most
species show temperatures which are roughly proportional to their masses (an observation with
barely any conjectures as to its understanding save that it may be related to either cyclotron pro-
cesses and/or shock processes). Minor ions also tend to flow somewhat faster than the main proton
component, by an amount which is roughly the (decreasing with distance) Alfvén speed.
Electrons - although the solar wind is generally regarded

p as a super-sonic flow, which can be re-phrased
as a flow faster than the thermal speeds ∼ kb T /m, this is not the case for the lighter electrons.
Electrons can travel freely out beyond the orbit of Jupiter and then return to the Sun. An inter-
planetary electric field is established to prevent the hot electrons running away.
5.5. EXERCISES 69
3-D and transient phenomena - we have some observations of the solar wind over the Sun’s poles
from the Ulysses spacecraft, and many observations of transient phenomena in the ecliptic plane.
While some 3-D time-dependent numerical models of the solar wind do exist, there remains much
work to be done.
5.5 Exercises
1. (a) Verify that the expression
V2 2kb T 4kb T GM
− lnV = ln r + +K (5.37)
2 m m r
is a solution to Equation 5.22 governing radial outflow of an isothermal solar wind.

(b) Use the conditions at the critical radius r = rc to determine K.
(c) Assume that the solution gives V ∼ constant for r ≥ (1/3)AU and that the solar wind ex-
pansion remains isothermal. Calculate the solar wind proton density, the magnetic field
strength, and the spiral angle at Mercury, Mars, Jupiter, and Neptune (at r =0.39AU, 1.5AU,
5.2AU, and 30AU respectively). Use the data given in Table 5.1 for the properties at Earth
to determine any constants you may need.
2. Theoretical estimates of the radial distance of the subsolar magnetopause from a magnetized
planet can be obtained by determining the distance at which the magnetic pressure B2 /2µ0 of the
planetary dipole field balances the sum of the solar wind ram pressure (ρV 2 ) and interplanetary
magnetic field pressure at the orbit of the planet. Using your data from the preceding question
on the properties of the solar wind and IMF at the 4 planets, and the table below, estimate the
stand-off distance of each magnetopause in units of planetary radius. Comment on the results.
[You will see that the interplanetary field has a negligible influence essentially because the solar
wind is super-Alfvénic.]
Planetary Radius (km) Magnetic Moment M (T m3 )

Mercury 2490 4.6 × 1012
Earth 6371 7.9 × 1015
Mars 3400 1.0 × 1012
Jupiter 71,400 1.5 × 1020
Neptune 24,800 2.0 × 1017
M
Note: Bsubsolar (r) ≡ 3
r
3. (a) Redo the derivation of Parker’s solar wind equation (5.22) but replacing the isothermal equa-
tion of state (5.10) by an adiabatic one
pρ−γ = C
(b) Find the adiabatic equivalent to the solution given in (5.37). Hint: Show that the equivalent
to the equation of motion (5.17) in this case can be written
γ
! !
d 1 2 d γ−1
d GM
V =− C ρ + (5.38)
dr 2 γ − 1 dr dr r
Chapter 6
Magnetic Reconnection
6.1 Introduction
The convection approximation (in the case of high magnetic Reynold’s number) introduced in Sec-
tion 4.3 leads to the picture of astrophysical plasmas as separate cells partitioned by thin current sheets
(see Figure 4.4), where, exactly because of their thinness, diffusion may be important. The local, small
scale breakdown of the Flux Freezing theorem will result in diffusion of field lines so that a field line
might now connect plasma elements from different sources. Thus the details of the very small scale
physics of magnetic diffusion, i.e., the physics of conductivity in a plasma with strong gradients, will
affect the global topology of the field line configuration.
If there is a strong field gradient such that the field on either side of the gradient is anti-parallel,
then diffusion of the field at the gradient can lead to a loss of total magnetic flux, and this is termed
magnetic annihilation. If on the other hand, there is an inflow into a limited diffusion region, then field
lines are continually being convected into the diffusion region and meeting field lines with an opposite
orientation. In this case the field lines can slip with respect to the plasma in the diffusion region, and
merge with field lines from the opposite side of the gradient. The “reconnected” field lines can then be
carried away from the diffusion region by the plasma outflow which balances the inflow. This situation
is known as magnetic reconnection. Magnetic reconnection is important because it allows the two sides
of the field gradient to be linked by a field line, and also because magnetic energy is continually liberated
in the process, causing accelerated and heated plasma flows.
Although everybody understands the importance of reconnection, it is fair to say that the micro-
physics of the diffusion process (i.e., the finite conductivity) is not well understood!
6.2 Magnetic Annihilation

6.2.1 Static Annihilation
Assume a non-flowing plasma boundary, and initially (at t = 0) a complete, discontinuous reversal of
the magnetic field direction, keeping constant field strength, so that there is a step function in Bx = ±Bo
>
for z < 0. The current sheet, i.e., where the field changes direction is the x − y plane. This configuration
is sketched in Figure 6.1.
The evolution of this problem is governed by the Induction Equation (4.13), with zero flow velocity:
∂B 1
= ∇2 B (6.1)
∂t µ0 σ
71
72 CHAPTER 6. MAGNETIC RECONNECTION
Figure 6.1: Geometry of a thin current sheet across which the magnetic field reverses over a scale 2`
This reduces, for our problem to the case of one spatial dimension:
∂Bx 1 ∂2 Bx
= (6.2)
∂t µ0 σ ∂z2
The solution of this partial differential equation, a diffusion equation, is
2 Z ξ −u2
Bx = B0 √ e du = B0 erf(ξ) (6.3)
π 0
where
!1/2
µ0 σ z
ξ= (6.4)
t 2
The current distribution corresponding

√ to the magnetic field given by (6.3) is Gaussian, and has a
width that increases in time as t (see Figure 6.2). Initially there is a rapid thickening of the current
sheet and a conversion of field energy to plasma energy (i.e., thermal energy), but this process becomes
slower and slower as the gradients weaken. Since the total flux in the system declines by the field lines
from either side diffusing towards the centre of the current sheet and disappearing at the magnetic null,
this is termed magnetic annihilation.
6.2.2 Dynamic Annihilation

In terms of converting magnetic energy to other forms (thermal, flow, etc.) static magnetic annihilation
is not particularly useful, because it switches itself off by dissipating the field gradients. Rapid flux
annihilation can be maintained if the plasmas on either side flow together so that the current sheet is
always squeezed, and the gradients maintained. In this case inward convection matches the expansion
from diffusion. A faster inflow leads to steeper gradients, which in turn leads to greater liberation of
magnetic energy.
We can estimate the thickness of the current sheet using two methods. In the first method we balance
the flux carried into the current sheet by the flow against the rate of flux annihilation.
Suppose the current sheet half width is `, and the speed of inflow (on both sides) is V . Then the flux
carried into the current sheet per unit time, per unit length (in y direction) is V B0 . On the other hand the
6.2. MAGNETIC ANNIHILATION 73
Figure 6.2: Evolution of an initially thin current sheet (left) due to magnetic diffusion. Note the broad-
ening of the current-carrying region and the accompanying reduction of magnetic field strength (and
hence energy).
rate of flux annihilation within the current layer can be estimated from the diffusion part of the induction
equation (e.g., (6.2) as
1 B0
` (6.5)
µ0 σ `2
Therefore, equating flux arrival and annihilation rates gives
1
`∼ (6.6)
µ0 σV
The other method to estimate the current sheet width is to note that both the inflow region (where
convection dominates) and the current sheet (where diffusion dominates) are associated with an electric
field, Ey in the y direction. For a steady state system which is independent of y (i.e., all quantities only
depend on (x, z)), one notes that Faraday’s Law reduces to ∇ × E = 0, so that Ey is spatially uniform,
i.e., the same in both inflow region and current sheet. So, in the inflow region Ey is just given by
the expression for the motional electric field: Ey = V B0 . In the diffusion region the electric field is
approximately given by the conductivity term in Ohm’s law, since the flow must be zero at the centre
of the current sheet: Ey ≈ jy /σ. The current jy in the y direction can be estimated from Ampere’s Law
(2.2)
B0
jy ≈ (6.7)
µ0 `
So that, by equating these two expressions for Ey , one again finds:
1
`∼ (6.8)
µ0 σV
We can now see that the magnetic Reynold’s number for the current layer is simply:
Rm = µ0 σV ` ∼ 1 (6.9)
As expected from the balance of magnetic diffusion and convection that we have assumed for the steady
current layer, the magnetic Reynold’s number for the layer itself is unity, rather than the much higher
value one would find for the overall system which sets up the current sheet.
In fact, we can find an analytic solution to the steady state induction equation, assuming equal and
<
oppositely directed flow Vz = ±Vi for z > 0, and assuming an undisturbed field of Bx = ±B0 far away
from the current sheet. Making the further, rather unsatisfactory, assumption that the flow continues to
be perpendicular to the current sheet and ∂/∂t = 0 the x-component of (4.13) becomes
d 1 d 2 Bx
0=− (Vz Bx ) + (6.10)
dz µ0 σ dz2
We shall solve this in the region z > 0 and then find the full solution using the symmetry of the problem.
If we further assume that the inflow is constant, i.e., Vz = −Vi we can then integrate once and re-arrange
to find
dBx
+ (µ0 σVi ) Bx = C (6.11)
dz
This is a first order linear ordinary differential equation with constant coefficients and can be solved
by the standard technique of a complementary function (satisfies RHS=0) and particular integral. The
general solution is
Bx = Ae−µ0 σVi z + D (6.12)
The constants A and D are determined by the boundary conditions that Bx → B0 as z → ∞ (which gives
D = B0 ) and Bx → 0 as z → 0 to provide continuity at z = 0. Hence
Bx = B0 1 − e−µ0 σVi z

z>0 (6.13)
The solution for z < 0 is found by noting from the symmetry of the problem that z → −z requires
Bx → −Bx , whence the full solution may be written
Bx = ±Bo 1 − e∓µ0 σVi z >

z< 0 (6.14)
Of course, the one dimensional “annihilation” geometries which we have discussed are completely
unrealistic since there is plasma inflow, but no exiting outflows, and we have assumed infinitely long
current sheets.
6.3 Models of Magnetic Reconnection

A more realistic picture is where the diffusion-dominated region is limited in space. The flows into these
“diffusion regions” are balanced by flows out along the current sheet.
A simple sketch of the main concepts is shown in Figure 6.3. In such a two-dimensional geometry
this would give a magnetic neutral line (where B = 0), and the field lines would have a so-called X-type
configuration. The place where the field lines “reconnect”, i.e., change their topology so that they link
the two sides of the current sheet is called the X-point. The reconnected field lines form stretched out
loops connecting the two sides. Magnetic tension in these loops accelerates the inflowing plasma away
from the neutral line. As the loops shorten they liberate energy which goes into heating the plasma. Thus
the outflow region is characterized by heated, accelerated plasma, and the convection term dominates in
this region. One expects the diffusion region to be very small. Nevertheless, this small region can affect
the global configuration of the magnetic field.
Note: we will be showing models of exactly anti-parallel reconnection, but reconnection can occur
in more general geometries, in which case one can consider the reconnection of just the anti-parallel
components of the field.
6.3. MODELS OF MAGNETIC RECONNECTION 75
Figure 6.3: Sketch of reconnection in two dimensions. Magnetic diffusion in a small region results in a
global reconfiguration of the magnetic topology, together with outflows along the current sheet.
6.3.1 Examples of Reconnection

Before we proceed to the more theoretical considerations, let’s have a look at some astrophysical phe-
nomena which are strongly linked with reconnection, and which have influenced greatly the development
of this subject.
Solar Flares
A model of rapid energy release in a solar flare is illustrated in Figure 6.4. A magnetic arcade or flux
rope can exist for an extended period in a MHD equilbrium in which the magnetic, plasma pressure, and
gravitational forces balance. This may include “line-tying”, in which the flux rope is held down by the
effects of magnetic tension of field lines overlying it, both ends of which map back to the surface of the
Sun, as illustrated in the top panel of Figure 6.4. In the pre-flare stage (middle panel) the injection of hot
plasma onto the flux rope from each end provides increased buoyancy and a slow rise of the structure
ensues. The overlying fields become more stressed, and energy is thus stored in this field. Finally, the
bottom panel indicates the rapid rise and release of this stored energy as a result of reconnection of
the stressed, closely anti-parallel field lines beneath the arcade. A second model in which reconnection
releases a “plasmoid” is shown in Figure 6.5.
Open and Closed Planetary Magnetospheres

Coupling between the solar wind and the Earth’s magnetosphere, both in terms of energy and plasma
particles, occurs as a result of reconnection between the interplanetary magnetic field and the terrestrial
magnetic field. When the IMF has a southward directed component of the field, reconnection occurs in
Figure 6.4: Illustration of a solar flare beginning (top to middle) with a slow rise of a line-tied flux rope
followed by a rapid release due to magnetic reconnection (see text).
the vicinity of the subsolar dayside magnetopause (between the field lines marked 1’-1’ and 1 in Fig-
ure 6.6). The magnetosphere becomes “open” and solar wind plasma is free to enter the magnetosphere
by travelling along the newly reconnected field lines. In addition, the magnetic tension of the newly
reconnected field lines (e.g. those labelled 2 and 2’ in the figure, which are sharply bent) accelerates
this plasma away from the reconnection site and over the poles of the Earth. These field lines eventually
straighten (3,3’), but the external portion remains frozen into the solar wind flow, and therefore con-
tinues to convect tailwards (4,4’). This results in highly stretched field lines (5,5’) being added to the
magnetic tail, which then slowly convect towards the tail centre plane due to the E × B drift related to
motional electric field E (E is everywhere out of the page in this 2-D representation, since it must be the
same everywhere from Faraday’s law and is directed out of the page in the solar wind region where it
can be readily evaluated). The magnetotail is therefore a reservoir in which energy extracted from the
solar wind flow is stored as magnetic energy. This also creates a central current sheet separating the
anti-parallel field in the northern and southern halves of the tail (the lobes). Reconnection also occurs
at this point (between the field lines marked 6,6’) creating both closed field lines (marked 7) and inter-
Figure 6.5: Illustration of a solar flare model in which magnetic reconnection releases an upward-moving
plasmoid while the downward flux of particles gives rise to the X-rays and other flare-associated phe-
nomena. (figure from Shibata et al., 1995) Note the similarity with the magnetospheric tail reconnection
phenomena illustrated in Figure 6.6.
planetary field lines (marked 7’). The tension on these field lines accelerates plasma both earthward and
tailward of the tail reconnection site, thus releasing the stored magnetic energy back to the plasma. The
Earthward convection of plasma and magnetic flux eventually returns the closed field lines (8) back to
the dayside (9), thus completing the magnetospheric convection cycle. Note that this flow of plasma
within the magnetosphere (jets over the poles, drifts towards the tail centre plane, accelerated jets away
from the tail reconnection site and then flows around the Earth at lower latitudes to return flux to the
dayside) are the result of a highly localized break-down of the frozen-in flux condition at 2 points - the
dayside magnetopause and the tail current sheet. Despite their limited extent (we have yet to directly
encounter one with spacecraft), they dramatically affect the global system - this convection pattern is
not possible without reconnection.
Figure 6.6: An open planetary magnetosphere. Plasma convection (and magnetic topology) is controlled
and driven by reconnection at the magnetopause and in the tail. The text provides a description of the
sequence of events numbered on the various field lines.
6.3.2 Sweet-Parker Model of Reconnection

Consider a time steady (∂/∂t ≡ 0), and two dimensional geometry (∂/∂y ≡ 0). Further assume that the
plasma is incompressible so that mass density ρ is constant. The current sheet has symmetric flow in
from both sides at speed Vi , and the plasma flows out along the current sheet at speed Vo , as sketched in
Figure 6.7. Using the same arguments as before, we expect the thickness of the current sheet (diffusion
region) to be
1
`≈ (6.15)
µ0 σVi
Suppose that ` stays constant for the system size L, where L is known. (That L is known and constant
turns out to be a crucial assumption.)
From conservation of mass we know that the mass entering the diffusion region per unit time (and
unit y-distance) is equal to that leaving out the ends, i.e.,
Vi L = Vo ` (6.16)
We will show that the outflow speed is approximately the Alfvén speed vAi of the inflow region.
Figure 6.7: Geometry for the Sweet-Parker model of reconnection.
We can estimate the outflow velocity based on conservation of energy. The Poynting flux is the flux
of electromagnetic energy, S = E × B/µ0 , and this is converted almost entirely into plasma energy. In
situations in which magnetic field energy is being converted to plasma kinetic and/or thermal energy we
find that j · E > 0, reflecting the dissipation of electromagnetic energy. In steady state one can express
energy conservation as
∇·S+j·E = 0 (6.17)
i.e., the electromagnetic energy flux diverges if there are net sinks or sources. But rather than work with
this equation we will simply balance the electromagnetic energy flux into the current sheet against the
kinetic energy of the outflowing plasma.
The electromagnetic energy into the current sheet per unit area per second is
Ei Bi Vi B2i
WE = = (6.18)
µ0 µ0
where we have used the convection electric field in the inflow region, Ei = Vi Bi . The mass of plasma
entering the current sheet, per unit area per second is
Fm = ρVi (6.19)
and this is accelerated to speed Vo . If we ignore the initial kinetic energy and any heating, the work done
on this plasma is
ρViVo2
WP = (6.20)
2
By balancing WE and WP we find

√ Bi √
Vo ≈ 2√ = 2vAi (6.21)
µ0 ρ
That is, as above, the outflow speed is approximately the Alfvén speed in the inflow region. This is
generally true independent of the reconnection rate (there are several measures of the “rate” of recon-
nection, but typically either the inflow velocity Vi suitably normalised or equivalently the electric field
Ey is used). The reconnection rate in this case determines how much plasma is accelerated, rather than
how fast.
Given the magnetic Reynold’s number for the inflow region: Rm = µ0 σvAi L, the conservation of
mass using our result (6.16) for Vo , and the width of the diffusion region ` from (6.15), we find
s
vAi 1
Sweet-Parker Reconnection Rate Vi ≈ = vAi √ (6.22)
µ0 σL Rm
[Using Vi instead of vAi in Rm actually makes the situation worse, as then the same calculation results
in Vi ∼ vAi /Rm . Additionally, since vAi is a characteristic speed of the medium, its use in Rm is model
independent.] And this poses a very serious problem, because the speed of inflow, which controls the
rate at which flux is brought into the diffusion region will be small if the magnetic Reynold’s number of
the inflow region is large, which will generally be true.
For example, in a solar flare, which we think is rapid energy release associated with magnetic recon-
nection, Rm ∼ 108 → 1010 , which with an Alfvén speed of 100 kms−1 and L ∼ 104 , would give a time
scale of tens of days! But we know that a solar flare releases its energy over the time scale of minutes
and hours.
6.3.3 Petschek Reconnection Model

The difficulty in the Sweet-Parker model is that the amount of material processed, and hence energy
liberated, is controlled by the width of the exit channel, which is controlled by the thickness 2` of the
diffusion region. The Petschek model circumvents this problem by widening the exit channel and having
standing shock waves (actually MHD slow-mode shocks) on the edges of the exit channel which perform
most of the conversion of magnetic energy to flow energy. (see Figure 6.8)
At the slow shocks the field and flow change abruptly in direction and strength: the field decreases
in the outflow region, and the magnetic energy change is balanced by an increase in the flow speed. In
the frame of the figure the shocks are stationary (standing shocks), but in the frame of
pthe plasma in the
inflow region they travel along the magnetic field at the inflow Alfvén speed vAi = Bi /(µ0 ρi ). Thus,
to ensure standing shocks we require:
Vi cos ξ = vAi sin(χ − ξ) (6.23)
where ξ is angle of shock to x-axis, and χ is angle of inflow magnetic field to x-axis. Here we have
equated the plasma inflow speed normal to the shock with the component normal to the shock front of
the shock propagation vector along the field (at an angle χ).
Next we note, by Faraday’s Law, that Ey is the same in the inflow and outflow regions:
Eiy = Vi Bi cos χ = Vo Bo = Eoy (6.24)
Then, from ∇ · B = 0 the normal component of the magnetic field on the two sides of the shock is the
same:
Bi sin(χ − ξ) = Bo cos ξ (6.25)
Figure 6.8: Geometry of the Petschek model for reconnection. Note the wide exit channels by compar-
ison to the fixed-width ones in the Sweet-Parker model shown in Figure 6.7. These exit channels are
bounded by slow shocks (dashed lines) which compress and heat the plasma while decompressing the
magnetic field.
Therefore
Bi cos ξ
Vo = Vi cos χ = Vi cos χ (6.26)
Bo sin(χ − ξ)
But sin(χ − ξ) = Vi cos ξ/vAi so
Vo = vAi cos χ (6.27)
In other words the outflow is roughly the same as the inflow Alfvén speed, reminiscent of our earlier
work in the Sweet-Parker model (see Equation 6.21).
Next, we conserve mass across the shock boundary:
ρiVi cos ξ = ρoVo sin ξ (6.28)
where ρo > ρi , i.e., there is compression at the shock (and since the magnetic field weakens this shock
is related to the slow magnetosonic mode - see Exercise 5 on p.57 - in that the magnetic and thermal
pressures are in opposition). Therefore one can find that:
ρi 1
Vo = Vi (6.29)
ρo tan ξ
Eliminating Vo between (6.27) and (6.29) and then eliminating the angle ξ by expanding sin(χ − ξ) =
sin χ cos ξ − cos χ sin ξ in (6.23) we reach
sin χ
Petschek Reconnection Rate Vi = vAi (6.30)
ρi
!
1+
ρo
Since 0 < ρi /ρo < 1 it follows that

1 Vi
< <1 (6.31)
2 vAi
In other words, the inflow speed, which determines the reconnection rate and the rate of energy lib-
eration, is now a reasonable fraction of the inflow Alfvén speed. More precise calculations show that
the fraction is about 0.1. The Petschek model provides the justification that magnetic reconnection is a
viable mechanism for energy release and magnetic topology reconfiguration in astrophysical plasmas.
6.4 Evidence for Reconnection

As discussed above, we do not understand the microphysics of the diffusion region, and they are ev-
idently small enough to have so far missed detection by spacecraft. However, observations of rapid
energy release in solar flares and the global convection pattern within the magnetosphere are strongly
suggestive that such a process must be occurring. In addition, models of reconnection suggest that as
a result of this process, plasma flows at speeds of order the inflow region Alfvén speed should be pro-
duced. These flows are large scale, and thus should be more easily detected. This is indeed the case, as
illustrated in Figure 6.9 which shows 2 hours of observational data from the ISEE-2 spacecraft, outbound
across the dayside magnetopause on August 12th 1978. From top to bottom, the 5 panels show the ion
number density, the ion temperature, the magnitude of the ion flow velocity, and the y-component of the
same, and the north-south (Bz ) component of the magnetic field, each as a function of time. At 1700
hours universal time (UT), the spacecraft is in the magnetosphere and observes field and plasma parame-
ters typical of this region - a low ion density, but high temperature, a strong positive Bz component of the
field, and very little plasma flow. This contrasts sharply with the data at 1900 UT, when the spacecraft
has passed out into the magnetosheath (shocked solar wind) where the plasma is relatively denser and
cooler, and the field strength is weaker (and in this case has a negative Bz component, antiparallel to
the magnetospheric field). The current sheet separating these two distinct plasma regions was crossed
just after 1830 UT, as designated by the two vertical lines, where there are strong gradients in the mag-
netic field. However, the important thing to note is that there are intervals prior to 1830 UT, when the
spacecraft is still in the magnetosphere (the magnetic field is unchanged), but the observed plasma has
densities and temperatures which are more characteristic of the external magnetosheath. This indicates
that the magnetosheath plasma has access to the magnetosphere, which it can only do if the frozen-in
flux condition has broken down globally (unlikely) or if the magnetospheric and magnetosheath fields
have become connected and the sheath plasma can stream in along the field lines. This, of course, is
possible if reconnection is occurring at a nearby diffusion region. Moreover, note that the plasma flow
speeds in these regions are higher than those in both the magnetosphere and magnetosheath. In other
words, this magnetosheath-like plasma has undergone acceleration as it crossed the magnetopause cur-
rent layer. Although the exact calculation needs to be done in the rest frame of the plasma transverse to
the magnetopause, it is possible to show that these flows are in fact of order the upstream Alfvén speed,
as predicted by all the reconnection models. This is strong evidence for the reconnection process being
of fundamental importance in astrophysical and space plasmas.
6.4. EVIDENCE FOR RECONNECTION 83
Figure 6.9: Data from the ISEE-2 satellite showing the ion and magnetic field variations in the vicinity
of the Earth’s magnetopause (see Figure 6.6). The text describes the evidence for accelerated flows and
plasma access consistent with magnetic reconnection occuring between the Earth’s magnetic field and
that of the solar wind in the magnetosheath region behind the Earth’s bow shock.
6.5 Exercises
1. For the magnetic annihilation solution given in Equation 6.14, find the electric field E (you
should remind/convince yourself that this electric field is everywhere constant). Hence calcu-
late the Poynting flux S = E × B/µ0 and show that it is directed toward z = 0. Show further that
|S(z → +∞)| is equal to 0∞ E · jdz and interpret this result.
R
2. Consider a region downstream from a reconnection neutral line in which the hairpin-like field lines
are given by B = (Bx (z), 0, Bz ) where Bx reverses across a current sheet at z = 0, Bz is constant,
and Bz Bx (i.e., the angle χ between the field and the current sheet is small such that cos χ ≈ 1).
A spatially uniform electric field E = (0, Ey , 0) exists across the system. Assume the plasma in
the system is cold, of number density n, and subject only to E × B drifts in the observer’s frame
of reference.
(a) In the frame in which the electric field is transformed away, the hairpin-like field lines are
stationary, there is no E × B drift, and particles move along the field line but cannot change
their speed. By balancing the magnetic tension outside of the current sheet with the change
of momentum of particles as they move through the current sheet and change direction, find
the speed of the particles in this frame.
(b) Hence show that in the rest frame of the observer
√ the field lines move along the current
√ sheet
away from the neutral line at a speed vA / 2 and that the outflow plasma speed is 2vA .
(c) Consider a line of unit length in the y-direction. In steady state, the magnetic reconnection
rate can be considered as equivalent to the amount of magnetic flux crossing this line per
unit time. Show that the reconnection rate is thus equivalent to Ey and that the reconnection
rate determines Bz in the system.
(d) An alternative way of visualising the acceleration of particles is as follows: In the field line
rest frame, particles move into the current sheet along the x-direction at the speed determined
in (2a). In the current sheet, their direction of motion is reversed by performing a half
gyration in the Bz field (as Bx = 0 at z = 0). What distance is the particle displaced in the
y-direction?
(e) In the observer’s rest frame, a displacement in the y-direction means that the particle has
moved along the electric field Ey , and will thus have gained energy. Using the results above,
relate the final kinetic energy to that gained√by moving along the electric field and hence
again show that the outflow plasma speed is 2vA .
Chapter 7
Shocks and Discontinuities
7.1 Shocks: Introduction
Note: most of the following was written for another purpose, and some time ago (!), so there is a
discontinuity in style.
The universe is woven through by plasmas in motion. Between the planets, between the stars,
between the galaxies there are flows of plasma and field energy, and wherever these flows exceed the
speed of sound and/or the Alfvén speed there will also be shock waves. In the solar system there
are shocks in front of all the planets, in their magnetotails, and formed in the solar corona and solar
wind. Shocks are the most studied nonlinear waves of plasmas, they are places where the plasma and
field go through dramatic changes: changes in density, temperature, field strength, flow speed. These
changes, combined with the collisionless nature of space plasmas and the wide variety of wave modes,
produce a rich collection of different shock types. In addition, shocks in collisionless plasmas are not
just interesting in their own right, but also because they are involved in a massive range of plasma
phenomena, which stretches and expands our knowledge of basic plasma processes.
Most of the everyday notions of what a shock wave is come from supersonic aircraft or explosive
blasts. The study of shocks began with ordinary gas dynamics in the late nineteenth century, and reached
its maturity during the 1940s, at the time of the development of high performance aircraft. The study
of plasma shocks surfaced during the 1950s, with interest in fusion plasmas and shocks caused by ex-
plosions in the upper atmosphere. But also, it was realized, shocks would exist in collisionless plasmas,
such as found in interplanetary space (the solar wind) and in other more exotic astrophysical objects.
Collisions in an ordinary gas serve to transfer momentum and energy among the molecules, and they
provide the coupling which allows the basic wave, the sound wave, to exist. In a collisionless plasma
the collisional coupling is absent. This is expressed by saying that the mean free path between collisions
is greater than the size of the system. For example, in the solar wind the collision mean free path, as
calculated from gas kinetic theory, is about 1AU (1AU=distance from Earth to sun = 1.5 × 108 km). But,
from observations, the thickness of the Earth’s bow shock, which forms in front of the magnetosphere, is
only 100–1000 km. This means that, whatever is happening at the shock, collisions cannot be important,
and instead there are processes in operation which are unique to collisionless plasmas. Since the start of
spaceflight we have collected many observations of shocks, to such an extent that laboratory collisionless
shock studies have been eclipsed. The ability to make detailed observations of the particle distributions
as well as the fields means that we have the opportunity to study in close up a phenomena which we
know exists in similar forms throughout the universe.
85
86 CHAPTER 7. SHOCKS AND DISCONTINUITIES
7.2 Shocks Without Collisions

The plasmas that are found in the magnetosphere, in interplanetary space and elsewhere in the universe
are very different from an ordinary gas. As we have seen in Chapters 4 and8 plasma can support several
different types of wave, involving fields and particles, rather than the single sound wave of a gas. But
the greatest difference is that most space plasmas are collisionless. This means that the media are either
so rarefied (i.e., under-dense) or hot, that Coulomb collisions between the constituent particles happen
so infrequently that they do not play an important role. In a normal gas collisions between molecules
ensure that they all have the same temperature, irrespective of type; collisions provide the mechanism to
propagate pressure and temperature changes; dissipation in the form of viscosity is also an outcome of
collisions. Collisions also ensure that in equilibrium the distribution of molecular speeds is Maxwellian.
Knowing what collisions do in an ordinary gas it is fairly easy to list what the lack of collisions
produces in a plasma. Different types of particles, e.g., protons and electrons, can have different temper-
atures. The particle distribution functions can be very different from Maxwellian, so that the concept of
temperature has to be broadened to that of “kinetic temperature,” i.e., a velocity moment of the particles’
distribution function. The important role of the magnetic and electric fields in a plasma also means that
the distribution functions may no longer be isotropic in velocity space. All these effects, and more, lead
to a superabundance of phenomena involving particles and fields.
Examples of collisionless shocks are spread throughout the universe. The most studied is the Earth’s
bow shock, which is a standing shock in the solar wind ahead of the Earth’s magnetosphere. The
magnetic field of the Earth forms an obstacle to the supersonically flowing solar wind. The bow shock
makes the solar wind transition to subsonic speeds, so that it can flow around the magnetosphere. It has
a curved shape, symmetrical about the Sun-Earth line, close to a paraboloid of revolution. The position
of the nose of the bow shock (the most sunward part) is at about 15RE from the Earth’s center (an Earth
radius = 1RE = 6371km). The shock is a few RE ahead of the magnetosphere, and this distance is called
the stand-off distance. The exact position of the bow shock relative to the Earth depends on the solar
wind ram pressure, because the magnetosphere is slightly “spongy.” The stand-off distance is a function
of the shape of the obstacle; blunt obstacles have larger stand-off distances. The region of subsonic solar
wind behind the bow shock is called the magnetosheath. Typically the interplanetary magnetic field (i.e.,
the solar magnetic field as carried by the solar wind) is at an angle of 45◦ to the Sun-Earth line. Thus,
the magnetic field intersects the shock at different angles around its curved surface, with corresponding
changes in the character of the shock. All of the planets with magnetospheres or ionospheres have bow
shocks in front of them, and most have been seen at least once or twice by interplanetary probes such as
Voyager, Pioneer, etc.
Shocks occur in the solar atmosphere (the corona) during solar flares and other manifestations of
solar activity. Flares and coronal mass ejections can inject energy and material into the solar wind
driving travelling interplanetary shocks which propagate out through the solar system. The solar wind
has high speed and low speed streams, coming from different source regions on the sun. Shocks can form
at the interface between a slow stream being overtaken by a fast stream. In more exotic astrophysical
bodies one finds jets of material from active galactic nuclei (AGN), and there are probably shocks formed
at the interface between the jet material and the interstellar medium. In supernovae massive amounts of
energy are deposited in a very short time, and shocks are formed as the supernova remnant (SNR) piles
up material as it expands away from the newly formed pulsar.
It would seem that lack of collisions could make our problem impossibly complex. However, to an
extent we are saved by the fact that the action of the magnetic field tends to replace the role of collisions
in “binding” the particles of the plasma together. So we will base our analysis on an MHD description of
the plasma. Since MHD does not include, except by trickery, effects due to individual particles (“kinetic”
effects), it cannot tell us anything about how a shock provides dissipation, or what the structure of the
shock will be. But MHD will be suitable to describe the plasma far upstream and downstream of the
7.3. SHOCK CONSERVATION RELATIONS 87
shock itself.
7.3 Shock Conservation Relations

However bizarre a shock may be we know that mass, energy and momentum will be conserved. We can
use MHD to relate the plasma states upstream and downstream of the shock using these conservation
relationships. Note that in this description the shock is very much a “black box,” which changes the
state of the plasma (field, density, etc.), while we can’t tell anything about what is actually happening
within the shock. Most importantly we don’t know anything about how “big” (i.e., thick) the shock is,
since MHD has no fundamental length scale, and the scales which should be important (e.g., particle
gyroradius) have been left out of MHD. In the case of an ordinary gas or fluid the relations between
the upstream and states were first derived by Rankine and Hugoniot towards the end of the nineteenth
century. In the case of a collisional gas these “Rankine-Hugoniot” relations determine uniquely the
downstream state in terms of the upstream state. This has the important result that in an ordinary gas
there is a unique transition between a given supersonic flow and a subsonic state. The shock structure is
determined by the dissipation mechanism, which is usually just viscosity. This result produces a shock
which has a thickness of just a few collisional mean free paths. Actually, even an ordinary gas can have
a more complicated shock structure, especially when the shock conditions produce chemical reactions
(e.g., disassociation of molecules), or partial ionization.
For a collisionless plasma, except for the single fluid case, the conservation relations (also known as
the shock jump or Rankine-Hugoniot relations) do not provide a unique prescription for the downstream
state, mainly because energy conservation only gives information about the total pressure (and hence
temperature), and not about how it is divided amongst the different types of particles in the plasma. In
other words, we need to know about the shock structure, about how the shock works, in order to know
how much the ions and electrons heat in passing through the shock.
In deriving the MHD Rankine-Hugoniot relations we make certain assumptions, such as that the
shock is, on average, stationary in the shock frame; that the energy in waves is not important; that the
particle distributions can be described by Maxwellians (or at least by low order moments such as the
density, bulk velocity, etc.) and the plasma pressure is isotropic. One of the basic observational tests
for a shock is a comparison with the Rankine-Hugoniot relations. This is often a basis for a discussion
about whether the observed “thing” is actually a shock. Often, what are really being tested are the
assumptions which go into our derivation of the jump conditions. In this case it is best to return to our
basic definitions of a shock which we discussed earlier. The Rankine-Hugoniot relations are only one
possible expression of the conservation of energy and momentum.
Let us consider the simple case of a one-dimensional, steady shock. We will work in a frame where
the shock is stationary. The x axis will be aligned with the shock normal, so that the plane of the shock
is parallel to the y-z plane. We will also assume uniform upstream and downstream media (flow, field,
density, and pressure) away from the shock transition itself. We can think of the shock as a discontinuity,
but in reality it will have some thickness given by the kinetic processes at the shock. The basic geometry
is shown in Figure 7.1. We shall see later that the reference frame and axes can be chosen in such a way
that the flow and field are confined to the x − z plane.
The shock separates two regions of steady flow. Plasma flows into the shock on one side (upstream)
and out the other side (downstream), and we label these two regions u and d, respectively. The shock
causes a change in the plasma description from mass density ρu , velocity Vu , magnetic field Bu , pressure
pu , etc., to the downstream values ρd , Vd , Bd , pd . The jump across the shock in any quantity X can be
expressed using the following notation:
[X] = Xu − Xd (7.1)
The MHD description gives us a set of conservation equations for the mass, energy, and momentum. For
Figure 7.1: Geometry and notation for the derivation of the 1D Rankine-Hugoniot relations for a shock
transition between the super-“sonic” upstream flow and sub-“sonic” downstream state. The velocity and
field vectors are not, ab initio, confined to the y − z plane both upstream and downstream.
any quantity a conservation equation has the form:
∂Q
+∇·F = 0 (7.2)
∂t
where Q and F are the density and flux, respectively, of that conserved quantity. If the shock is steady
(∂/∂t ≡ 0) and one dimensional (ie., there are variations only along the x axis, so that ∂/∂y ≡ ∂/∂z ≡ 0),
then (7.2) implies that
d
(Fx ) = 0 (7.3)
dx
which in turn implies that (Fu − Fd ) · n̂ = 0, where we denote the unit vector normal to the shock surface
by n̂. Therefore, the component of the conserved flux normal to the shock remains constant, and this can
be written:
[Fn ] = 0 (7.4)
The subscript n indicates the normal component, which, for the geometry we are considering, is the
same as the x component.
For MHD we have the equation for the conservation of mass, or continuity equation (4.1), which is
already in the form (7.2). In 1-D we thus have
d
(ρVx ) = 0 (7.5)
dx
which leads to the jump condition for the shock:
[ρVx ] = 0 (7.6)
This tells us, as we would expect, that if the shock slows the plasma, then the plasma density increases.
The MHD momentum equation (4.2) and equation of state (4.3) are not in conservation form. The
momentum equation can be re-cast, using both the continuity equation and ∇ · B = 0, into the following
conservation form
∂(ρV) B2

BB
+ ∇ · ρVV + p + II − =0 (7.7)
∂t 2µ0 µ0
Dotting this equation with V and making use of all the other MHD equations leads, somewhat tediously,
to a total energy conservation equation
∂ 1 2 B2 γ

p 1 2 E×B
ρV + + + ∇ · V ρV + pV + =0 (7.8)
∂t 2 γ − 1 2µ0 2 γ−1 µ0
These allow us to write down the corresponding jump conditions. Firstly, the conservation of mo-
mentum normal to the shock surface is
B2

ρVx + p +
2
=0 (7.9)
2µ0
This reflects the momentum changes due to the pressure and magnetic forces, although in this form it is
clear that the pressure and magnetic field can be thought of as carrying momentum fluxes in their own
rights.
The transverse momentum also has to balance, and this gives:

Bx
ρVx Vt − Bt = 0 (7.10)
µ0
The t subscript indicates the components transverse to the shock, i.e., parallel to the shock surface. This
equation reflects the tangential stresses related to any bend or kink in the magnetic field (since ρVx is
constant from (7.6) and we shall see below that Bx is also constant).
The final conservation equation from MHD is for the energy. We have assumed that the plasma has
an adiabatic equation of state so that pρ−γ = constant. For a normal monatomic gas γ = 5/3. Actually,
the real equation of state will probably not be adiabatic, but the results we obtain will still be qualitatively
correct. The shock jump condition from the energy conservation Equation 7.8 is:
γ p B2

1 2 Bn
ρVx V + +Vx − V · B =0 (7.11)
2 γ−1 ρ µ0 µ0
The first two terms are the flux of kinetic energy flow energy and internal energy. The last two terms
come from the electromagnetic energy flux E × B/µ0 where we have used the ideal MHD result that
E = −V × B.
Equations 7.6, 7.9, 7.10, and 7.11 are the jump conditions for the gas, but there are also purely
electromagnetic boundary conditions. From Maxwell’s equation ∇ · B = 0 the normal component of the
magnetic field is continuous (Bx = constant):
[Bx ] = 0 (7.12)
∂B ∂
From Faraday’s Law ∇ × E = − , with the assumption that = 0, the tangential component of the
∂t ∂t
electric field must also be continuous. Using E = −V × B, this becomes:
[Vx Bt − Bx Vt ] = 0 (7.13)
The set of conservation relations have been found with the intention of using them to calculate shock
jump conditions, but in fact we have not explicitly forced the solutions of Equations 7.6, 7.9, 7.10,
Table 7.1: Possible types of discontinuity/shock in ideal MHD
Discontinuities
Density jump arbitrary, but pressure and
Contact Discontinuity Vu = 0, Bn 6= 0
all other quantities are continuous.
Plasma pressure and field change main-
Tangential Discontinuity Vx = 0, Bn = 0
taining static pressure balance.
Form of intermediate shock in isotropic
√
Rotational Discontinuity Vn = Bn / µ0 ρ plasma, field and flow change direction
but not magnitude.
Shock Waves: Vu 6= 0
Parallel Shock Bt = 0 Magnetic field unchanged by shock.
Plasma pressure and field strength in-
Perpendicular Shock Bn = 0
creases at shock.
Oblique Shocks Bt 6= 0, Bn 6= 0
Plasma pressure and field strength in-
Fast Shock crease at shock, magnetic field bends
away from normal.
Plasma pressure increases, magnetic field
Slow Shock strength decreases, magnetic field bends
towards normal.
Intermediate Shock Only shocklike in anisotropic plasma.
7.11, 7.12, and 7.13 to be shock-like. The solutions of these equations describe a number of different
types of MHD discontinuity, including shocks. For a discontinuity to be a shock there must be a flow
of plasma through the shock surface (Vn 6= 0), and there must be some dissipation and compression
across the shock. A further distinction can be made between discontinuities which are threaded by a
magnetic field (i.e., Bn 6= 0), and those which are not. Shocks with a normal magnetic field component
are called oblique, which refers to the angle between the shock normal and the upstream magnetic field
(see below). Table 7.1 summarizes the usual classification of MHD discontinuities. We shall give some
specific examples of the use of the conservation relations below, but there are a few general points to be
made first.
The conservation relations are a set of six equations. If we wish to find the downstream state in
terms of the upstream state, then there are six unknowns: ρ, Vn , Vt , p, Bn and Bt . This means that the
downstream state is specified uniquely by the conservation equations, as in ordinary fluid theory. How-
ever, we only have to introduce either an anisotropic pressure (e.g., pressures parallel and perpendicular
to the magnetic field are often used), or another fluid (e.g., electrons or heavy ions), and then there will
be more unknowns that equations. In such cases we are forced to use additional relations, from theory
or observation, to provide the missing information.
As seen in Table 7.1 the oblique shocks are divided into three categories, the fast, slow and interme-
diate (sometimes known as Alfvén), which corresponds to the three modes of small amplitude waves in
MHD. The intermediate shock is really a special case, since it is only shock-like under special circum-
Figure 7.2: Refraction of magnetic field and flow away from (left) or toward (right) in the case of a fast
shock (left) and slow shock (right).
stances. In an isotropic plasma (as we have been dealing with) it is not a shock, and is rightfully called
a rotational discontinuity; it will not be discussed further.
The fast and slow shocks have the same behavior of plasma pressure and magnetic field strength as
at the corresponding MHD linear waves derived in Section 4.6.1, but the shock jump conditions are fully
nonlinear. This can be explained by observing that the shock jump conditions are valid for shocks of any
strength. In particular, they must be true for very weak shocks. For consistency then, in the weak shock
limit the shock jump relations must correspond to the modes of small amplitude MHD waves. Thus
even in the fully nonlinear (large amplitude limit) the shock relations have the heritage of the linear
modes. Another explanation is that the formation of a shock can be produced by the steepening of a
large amplitude wave. Such a wave could steepen because the speed of waves with shorter wavelength
could be changed by the wave itself. This would produce a distortion of the wave, with some parts
of the wave moving relative to others. This could eventually lead to sharp gradients, and hence shock
formation. Such shocks would again be expected to retain characteristics of the original mode of the
wave.
Across a fast mode shock the field strength increases, but the normal component is constant, so that
the increase is all in the transverse component. Therefore, at a fast shock the downstream field turns
away from the shock normal. Conversely, at a slow shock the downstream field bends towards the shock
normal. These characteristics are illustrated in Figure 7.2. It seems from observations that fast shocks
are by far the most frequent type of shock observed in solar system plasmas. Planetary bow shocks are
fast mode shocks, as are most interplanetary shocks in the solar wind. Most of our discussion later will
concentrate on fast shocks. Slow shocks are rarer, although they have been observed, and they play an
important part in Petschek’s model of magnetic reconnection (see Figure 6.8).
7.3.1 The Exactly Parallel Shock
The parallel shock has the upstream magnetic field parallel to the shock normal, i.e., Bu = Bx n̂, But = 0.
We will use the conservation relations in the frame where the upstream flow is also parallel to the shock
normal, so that Vu = Vx n̂, Vut = 0.
An important result can be obtained by eliminating Vt from Equations 7.10 and 7.13 (and assuming
Vn 6= 0, Bn 6= 0). " ! #
B2n
1− Vn Bt = 0 (7.14)
µ0 ρVn2
In the case of a parallel shock the quantity in parantheses is in general non-zero. If But = 0, then to
satisfy (7.14) Bdt is also zero. Thus the direction of the field is unchanged by the shock. But since Bx is
the only non-zero component of the field, and since it does not change, it follows that the total magnetic
field is also left unchanged by the shock. There is a compression in the plasma, but not in the field.
Feeding this result back into the conservation relations removes all mention of the magnetic field. From
the MHD perspective this means that the shock is like an ordinary fluid shock, and the magnetic field
does not play a role. However, in the context of a collisionless plasma the only way for dissipation to
occur is via field-particle processes, so it is certain that here the fields somehow play a crucial role, e.g.,
in some wave/turbulent manner.
7.3.2 The Exactly Perpendicular Shock

At the exactly perpendicular shock the upstream field is perpendicular to the shock normal. In this case,
Bx = 0 and Bu = But . Again we examine the case where the upstream flow is parallel to the shock
normal: Vu = Vux x̂. To ensure shocklike solutions there will be a nonzero mass flux through the shock:
ρuVux = ρd Vdx 6= 0 (7.15)
We define a density compression ratio r = ρd /ρu . Using the above equation we can write Vdx = (1/r)Vux .
We can now begin to apply the jump conditions. Equation 7.10 becomes
ρuVux Vut − ρd Vdx Vdt = 0 (7.16)
which implies that Vdt = 0, because Vut = 0 and ρVx 6= 0.
From Equation 7.13, using Bx = 0, the jump condition becomes:
Vux But = Vdx Bdt (7.17)
This tells us that the upstream and downstream fields are parallel. Because there is no normal magnetic
field or transverse flow velocity, throughout the system, from now on we will simply use Bu , Bd , and Vu ,
Vd . Using the compression ratio r, we see that Bd = rBu . In other words, the field compresses as much
as the flow.
In the perpendicular case Equation 7.9 reduces to:
B2u B2
ρuVu2 + pu + = ρd Vd2 + pd + d (7.18)
2µ0 2µ0
which can be rewritten as
B2

1
ρuVu2 + (pu − pd ) + u 1 − r2 = 0

1− (7.19)
r 2µ0
After substituting for Vd and Bd , the energy jump condition equation (7.11) in the perpendicular
case, becomes:
γ B2

1 1 1
ρuVu 1 − 2 +
2
(pu − pd ) + u (1 − r) = 0 (7.20)
2 r γ−1 r µ0
Equations 7.19 and 7.20 can be used to eliminate pd , and we are left with an equation for r, the
compression ratio, in terms of the upstream parameters.
2 (2 − γ) γ

2
(r − 1) r +r + 2 + γ − 1 − (γ + 1) = 0 (7.21)
MA2 MA2 Mcs
In this equation we have introduced the Alfvénic Mach number MA , which is the ratio of the upstream
flow speed (along the shock normal) to the upstream Alfvén speed; similarly we define the sonic Mach
number Mcs as the ratio of the upstream flow speed to the upstream sound speed.
Vu
Alfvén Mach Number MA = √ (7.22)
Bu / µ0 ρu
and
Vu
Sonic Mach Number Mcs = p (7.23)
γpu /ρu
One of the solutions of equation 7.21 is clearly r = 1, which represents a downstream field, velocity,
and density unchanged from the upstream values. Obviously, this doesn’t correspond to a compressive
shock (although it is a good check on the algebra!). The quadratic term in (7.21) leaves two other
solutions, one of which is negative if γ < 2. This negative r solution is unphysical, and so we are left
with one solution for the compression ratio. An interesting limit is the high Mach number limit, when
MA 1 and Mcs 1. In this case Equation 7.21 becomes r(γ − 1) − (γ + 1) = 0, or:
γ+1
r= (7.24)
γ−1
Therefore there is a finite limit to the compression which depends only on γ. Remember that γ is just
an indication of how the plasma heats, and is not dependent on the upstream parameters. In particular,
for a monatomic gas γ = 5/3, and this gives a limiting compression of four. So at a high Mach shock
the maximum jump that can be expected in the field, density and velocity is by a factor four. But don’t
forget that this factor four, which is much quoted, depends on the details of how the plasma heats at the
shock, and a polytropic equation of state hides all these details in an effective (assumed constant) γ.
Another consequence of the high Mach number limit (i.e., large upstream flow energy) can be ob-
tained from equation 7.20 where the terms dependent on Bu and pu can be neglected. Remembering that
r is independent of the upstream parameters, there is thus a direct proportionality between the ram energy
of the flow 21 ρVu2 and the downstream thermal pressure pd . This is just an illustration of the operation
of the high Mach number shock: it takes the flow energy upstream, and converts it to thermal energy
downstream. Finally, at the perpendicular shock both the field strength and the density increase. From
the discussion above it follows that the shock is a fast mode shock. Indeed, it is possible to show from
(7.21), after noting that r > 1, that the upstream flow Vu must be faster than the fast magnetosonic speed
at perpendicular propagation (i.e., (4.58) in the upstream medium. Similar logic leads to the conclusion
that the downstream flow is slower than the downstream fast magnetosonic speed, so that the shock
does indeed bring about a transition from super-magnetosonic to sub-magnetosonic (see Exercise 7.5.2).
There is no slow mode perpendicular shock solution to (7.21). This is unsurprising given the result that
slow mode waves do not propagate perpendicular to the magnetic field.
7.3.3 Oblique Shocks

The general case of oblique shocks, i.e., those for which the upstream field is neither exactly parallel nor
perpendicular to the shock normal, is conceptually the same as those we have been looking at, although
the algebra is a bit more tedious. We shall see later that the angle θBn which the upstream field makes
with the shock normal n̂ plays a crucial controlling role in the processes occuring at such shocks. While
we will not perform the full Rankine-Hugoniot analysis in the general case, we note here that, as depicted
in Figure 7.2 and discussed in Table 7.1 both slow mode and fast mode transitions are possible.
One important feature of oblique MHD shocks emerges from Equation 7.14. Let us begin by noting
that it is always possible to move to a frame of reference in which the shock is at rest and the upstream
velocity is directed into the shock along the shock normal. This is called the “Normal Incidence Frame,”
or NIF. [Imagine the 3D version of Figure 7.1 in which the shock is at rest. Now translate along the shock
surface; the normal component of velocity doesn’t change, but you can make the tangential one take on
any value, including zero.] So in the NIF, there are only two directions of relevance to the upstream
region: the shock normal n̂ and the magnetic field Bu . These two vectors define a plane. Now (7.14)
tells us that the downstream tangential magnetic field Bdt is parallel to the upstream one, and hence the
total downstream magnetic field Bd lies in this same plane. Finally, the transverse momentum equation
(7.10) shows that any downstream tangential flow Vdt must be parallel to the tangential magnetic field,
so the downstream flow also lies in this same frame. If we started from a frame other than the NIF, we
would have concluded that the jump in tangential velocity [Vt ] would be confined to this plane, as is
a jump in the normal velocity component. This simply reflects the fact that the tangential velocity is
related to the field tension, which acts in the same direction as the “kink” in the field; there are no forces
directed out of this plane. Thus we reach
Coplanarity Theorem: At oblique MHD shocks, the shock normal n̂,

upstream and downstream magnetic fields Bu and Bd , and jump in veloc-
ity Vu − Vd are all coplanar.
The Coplanarity Theorem simplifies the analysis. It also makes it possible to estimate the direction
of shock propagation from real data since, for example, Bu × Bd and Bu − Bd are both perpendicular to
the normal, whence
(Bu × Bd ) × (Bu − Bd )
n̂ = (7.25)
|(Bu × Bd ) × (Bu − Bd )|
is the “magnetic coplanarity normal.”
There is one important exception to the coplanarity theorem, and that arises from inspection of (7.14)
when the expression in parentheses is zero, i.e.,
Bn
Vn = √ ≡ vAn (7.26)
µ0 ρ
or the normal flow is Alfvénic. In this case Bt can rotate out of the coplanarity plane. This special
case corresponds to a rotational discontinuity, and the other Rankine Hugoniot relations can be used to
demonstrate, in MHD, that there is no compression of the plasma, no deceleration, no heating, and no
change in the field magnitude. Thus rotational discontinuities are not shocks.
For completeness, there are also special cases when the tangential field is present on only one side
or the other. These are known as “switch-on” or “switch-off” shocks. These cases obey the coplanarity
theorem (on the side with no tangential field, there is only one defining direction, namely n̂).
7.4 Shock Structure

7.4.1 Real Shocks
So far we have described some fundamental concepts about shocks, and mentioned some examples of
where shocks will be formed. But in the case of shocks in space we are in an almost unique position,
7.4. SHOCK STRUCTURE 95
Figure 7.3: Observations of a supercricital quasi-perpendicular shock in space
since we have direct observations of naturally occurring shocks. It is possible to generate collisionless
shocks in the laboratory, and shock heating was one of the hopes for controlled nuclear fusion. But lab-
oratory experiments cannot approach the scale or global nature of the naturally occurring space shocks.
Space observations are also unique in that the smallest plasma scale (the Debye length) is usually larger
than the spacecraft. This means that we can truly make a point measurement, since our measurement
devices do not affect the plasma (at least if we are careful). With increasing space technology we are
making higher and higher resolution measurements of the space shocks. Not only are we measuring the
upstream and downstream states, but we can also measure how the plasma changes as it passes through
the shock. In other words, we can study the collisionless dissipation mechanisms in action.
Figure 7.3 shows the measurements made as a spacecraft passed through the Earth’s bow shock.
Instruments onboard measure the magnetic field. By counting particles arriving at the satellite, the
electron and proton distribution functions can be measured, and from these the density and the flow
speed of the solar wind as it passes through the shock can be calculated. The satellite has a very small
speed, and usually observations of the bow shock happen when the shock moves across the satellite.
This can happen because of slight changes in the solar wind speed or density. The observations consist
of time series for the different quantities, and if the speed of the shock relative to the satellite is constant,
then the profiles we observe in time will be the same as the shock’s profile in space. In the figure the
satellite is initially in the solar wind, and then the shock moves outwards and the observations are taken
as the shock moves over the satellite.
In passing from upstream to downstream it is obvious that the velocity decreases, and the density
increases. That is, there is compression at the shock. Also the magnetic field is shown, and this increases
like the density, so it is a fast mode shock. The next thing to notice is that although, the shock is thin, it is
not just a smooth transition. Instead, there is structure within the transition: There is a “foot,” a “ramp,”
and an “overshoot,” and we will see later that these are controlled by the way in which the solar wind
protons heat at the shock. For example, the thickness of the foot equals the distance to which protons
Figure 7.4: Shock geometry θBn (left) and the motion of reflected ions in the case of quasi-parallel
geometries (middle) and quasi-perpendicular ones (right).
are reflected by the shock before drifting through it.
7.4.2 Different Parameters Make Different Shocks

The most important result from space observations of shocks is that there are many different types of
shock depending on the shock parameters. We have already seen that the Rankine-Hugoniot relations
lead to different modes of shocks, but even restricting ourselves to fast mode shocks, like planetary
bow shocks, we find different types of shock structure. Early on there were quite serious debates about
whether the Earth’s bow shock was actually stable; perhaps all the different profiles seen in observations
were just fleeting glimpses of an ever changing entity. A major contribution was to show that there was
a definite pattern to the observations, and that the type of shock was determined by the complete set of
shock parameters describing the upstream conditions. The most important factors in controlling the type
of shock are the direction of the upstream magnetic field (relative to the shock surface) and the strength
of the shock.
We will first look at the influence of the direction of the upstream field, sometimes called the shock
geometry. Figure 7.4 shows a one-dimensional shock, and marked on it is the unit vector n̂ normal to
the shock surface. A convenient way to describe the direction of the upstream field is the angle θBn
between it and the shock normal. Depending on θBn the shock has dramatically different behaviors.
When θBn = 0 the shock is called parallel, and when θBn = 90◦ it is called perpendicular. The terms
quasi-parallel and quasi-perpendicular are used to divide the range of possible θBn values, with the
actual dividing line usually chosen as 45◦ . Another term, oblique, refers to a shock which is neither
exactly perpendicular or parallel, but it is also sometimes used to refer to the no man’s land between
quasi-parallel and quasi-perpendicular.
The importance of the parallel/perpendicular distinction is clear when we consider the motion of a
particle which is initially headed away from the shock. Such particles arise either by escaping upstream
from the downstream side or by being reflected at the shock itself. The motion of these particles is
particularly simple if we transform away the motional electric field by moving along the shock surface
7.4. SHOCK STRUCTURE 97
(towards the top of Figure 7.4) so that the upstream velocity Vu is aligned with the magnetic field Bu . In
this frame, the “de Hoffmann-Teller frame,” the shock is still at rest and now the particle motion is pure
gyromotion plus field-aligned guiding centre motion, since E0 = −V0 × B0 ≡ 0. It suffices, therefore, to
decompose the initial velocity of the particle into these components in order to follow these particles.
In the case of a quasi-parallel shock the field lines pass through the shock and a particle’s motion
along the field will carry the particle through (and away from) the shock relatively easily. On the other
hand, at quasi-perpendicular shocks the field lines are almost parallel to the shock surface, so particle
motion along the magnetic field would need to be very rapid (more technically, the pitch angle would
need to be quite small) if the particle is to escape from the shock. For most pitch angles, the particle
gyration at a quasi-perpendicular shock brings the particle back to the shock. There is a further simple
conclusion to be drawn from the parallel/perpendicular distinction: Because particle motion in the nor-
mal direction is “easier” at the parallel shock, compared to the perpendicular shock, then we can expect
that the scale of the parallel shock will be larger than for the perpendicular shock. Of course, this can
only be said with certainty once we know the exact dissipation mechanisms. Also our arguments about
particle motion are really only true for the ions, which have much larger gyroradii than the electrons, but
that is not so bad because they determine most of the structure of collisionless shocks. (Because of their
much greater mass the ions carry most of the mass and energy flow in a plasma.)
The shock angle θBn is the most important factor in controlling the shock type, but almost every
plasma parameter can have an effect: temperature, composition (i.e, what types of ions are present),
and shock Mach number M. The Mach number indicates the strength of the shock, and is a measure
of the amount of energy being processed by the shock. As might be expected, the higher the Mach
number the more dramatic the behavior of the shock. In the solar system shocks can be found with
Mach number between (almost) 1 and perhaps 20, but in more violent astrophysical objects, like the
shocks produced by super-nova remnants, the Mach number could be of the order of a thousand. The
processes operating at such shocks remain unclear, but for the solar system shocks we believe that we
have a fair understanding. However, even for solar system shocks this statement is only true for quasi-
perpendicular shocks. At the present time, quasi-parallel shocks are a topic of ongoing debate; the easy
particle access results in an extended region of waves, turbulence, and shock-energised suprathermal
particles.
In the quasi-perpendicular shock there is a clear distinction, at least theoretically, between a type
of low Mach number shock, the subcritical shock, and a higher Mach number shock, the supercritical
shock. For most of the time the Earth’s bow shock is supercritical, and to find subcritical shocks we
usually have to wait for a suitable interplanetary shock. As the name suggests there is a “critical” Mach
number which separates the two types of shock. This critical Mach number is at MA ∼ 2.7 for θBn = 90◦ ,
and decreases as θBn decreases. The Earth’s bow shock has values for MA in the range ∼ 1.5 − 10.
Observationally the difference between these two types of shock is clear. Our original example
(Figure 7.3) was of a supercritical shock. We have already mentioned the structure visible within the
shock: The field has a single sharp jump (called the ramp), but it is preceded by a gradual rise called
the foot which is related directly to the gyrating orbits of ions reflected by the shock and returned to the
shock surface as sketched in the right panel of Figure 7.4. Also, the field right at and behind the ramp is
higher than its eventual downstream value, and this is called the overshoot. Like good biologists we have
arrived at this separation into component parts only after seeing many examples. And we have ignored
what we don’t think is important, which in this case is all the small wiggles in the field. We label this
as small amplitude turbulence, which may play a role, but which doesn’t control or overwhelm the basic
shock structure.
In contrast with the structure of the supercritical shock, the subcritical shock resembles much more
the profile of a shock in an ordinary gas. Figure 7.5 shows several examples of low Mach number
shocks ranging from subcritical to slightly supercritical. The top two have relatively simple ramps from
the upstream to the downstream value, with little or no foot or overshoot. There is wave which appears
Figure 7.5: Observed profiles of the magnetic field magnitude at several quasi-perpendicular shocks
ranging from sub-critical (top) through to super-critical (bottom).
in the upstream region. This wave, the precursor, has a single, well-defined wavelength. The presence
of this precursor depends on both the θBn and the Mach number; it corresponds to non-MHD aspects
of shorter wavelength waves which travel faster than the fast magnetosonic wave and hence can escape
upstream. Notice that the wavelength is smaller in the second, more perpendicular, of these examples.
As supercriticality is reach and surpassed, the foot and overshoot become apparent. Here, too, the
size, shape, and extent of these features depend on the shock geometry and Mach number.
7.5 Exercises
1. Fill in the steps in Section 7.3.2 leading to Equation 7.21.
2. Shocks must be compressive, so the density ratio ρd /ρu ≡ r > 1. Re-arrange Equation 7.21 to
form an inequality making use of r > 1 and hence show that in this case of perpendicular shocks,
7.5. EXERCISES 99
q 1 ≤ γ ≤ 2, the upstream flow speed must be super-fast-magnetosonic, i.e., faster than

provided
v f = c2s + vA2 given in (4.58). For bonus marks, given this result construct a two-line (maximum)
argument to show that the downstream flow must be sub-fast-magnetosonic. [Hint: think about
the Rankine-Hugoniot relations in reverse.]
3. A planar slow mode shock is propagating through a plasma which, away from the shock itself,
obeys the ideal MHD equations. In the rest frame of the shock, the shock lies in the y−z plane, and
upstream of the shock the inflow velocity is Vu = −Vu x̂ where x̂ is the shock normal which points
into the upstream region. The upstream medium is uniform with mass density ρu , pressure pu ,
and magnetic field Bu = Bu (cosθu , 0, sin θu ). If the shock “switches off” the transverse magnetic
field component, so that in the downstream region Bd = (Bd , 0, 0), combine the continuity (mass
conservation) and z momentum jump conditions to show that the downstream region has a z-flow
component (the “jet” in Petschek reconnection) given by
Vdz cos θu sin θu

=
Vu MA2
where
Vu
MA ≡ √
Bu / µ0 ρu
is the upstream Alfvénic Mach number.
4. Referring to Figure 6.8 for the geometry of the slow mode shocks in Petschek reconnection, ex-
press the upstream shock conditions (ρ, V, B) in terms of the field geometry χ, shock geometry ξ,
electric field E and distant field strength B in a standard shock coordinate system. Assume that the
incoming plasma is cold, i.e., pu = 0. Use the Rankine Hugoniot relations to find the downstream
state Bd , Vd , ρd , and pd . Assume that the downstream field and flow directions are as shown in the
figure. [This question requires careful thinking and working.]
5. Sudden changes are detected in the solar wind and interplanetary magnetic field. The radial ve-
locity of the solar wind remains constant, but the density jumps from 5 to 10 cm−3 . The proton
temperature jumps from 5eV before the discontinuity to 13.8eV afterward, but the electron tem-
perature remains constant at 15eV. The magnetic field of (0, −8, 6)nT before the discontinuity
rotates and drops in strength to (0, 3, 4)nT. What type of discontinuity might this be, and why?
6. An interplanetary shock crosses the spacecraft Space Physics Explorer and its magnetometer de-
tects an upstream magnetic field of (6.36, −4.72, 0.83)nT and a downstream magnetic field of
(10.25, −9.38, 1.74)nT. Using the magnetic coplanarity theorem (and if needed [Bn ] = 0) de-
termine the orientation of the normal. The plasma analyzer detects an upstream velocity of
(−378, 33.1, 19.9)km/s and a downstream velocity of (−416.8, 7.3, 51.2)km/s. If the upstream
density was 7.5 protons/cm3 and the downstream density was 11 protons/cm3 , calculate the veloc-
ity of the shock. [Hint: In the rest frame of the shock, V∗u = Vu −Vsh n̂, V∗d = Vd −Vsh n̂, and you
will need to use the Rankine-Hugoniot relations. Assume that the solar wind is purely protons,
together with an equal number of electrons.]
Chapter 8
Multi-Fluid Models
8.1 Formulation
We have seen that the one-fluid MHD model of astrophysical plasmas, developed in Chapter 4 and
explored further in subsequent chapters, embraces many fundamental aspects. These include a com-
plete mathematical representation of mass conservation, fluid forces including electromagnetic effects,
an equation of state (with an ad hoc polytropic index), and the consequences of Maxwell’s equations,
especially magnetic induction, flux freezing, and diffusion. We left out, or glossed over, the implica-
tions of plasmas necessarily being comprised of more than one species (i.e., they must have at least
an ion species and electrons), phenomena on particle timescales (cyclotron and/or plasma frequencies),
transport processes such as conduction and viscosity, and interactions leading to instabilities.
Here we begin to address many of these deficiences by considering a plasma to be made up of
more than one species. To keep the problem manageable, we continue to work with fluid concepts as
opposed to the richer kinetic descriptions. Thus we assume that the plasma can be well-represented by
its fluid parameters, e.g., density, bulk velocity, temperature, etc., although we now each species can
have different values of these quantities.
Specifically, we denote the different species by the subscript “j.” Then we introduce the following
notation:
mj mass of a particle (8.1)

qj charge on a particle (8.2)
nj number density (8.3)
Vj bulk velocity (8.4)
pj pressure in species rest frame (8.5)
Note in particular that we have not defined a single, common centre of mass frame and measured all the
pressures in that frame. We treat each species on its own. While this has the advantage that it allows
us to describe each species better, it complicates some problems because the pressures are all measured
in different frames. In the centre of mass frame there are then contributions to the pressure which arise
from the relative bulk motions.
Now let us build the mathematical description of such a multi-fluid plasma. Most of this relies, as we
did in Chapter 4, on concepts of conservation equations and fluid elements. For example, conservation
of the number of particles of each species requires
∂n j
+ ∇ · (V j n j ) = 0 (8.6)
∂t
101
102 CHAPTER 8. MULTI-FLUID MODELS
by comparison with (4.1). This is usually referred to as the continuity equation.

The equation of motion for a fluid element of species “j” is (cf., (4.2)
∂
!
+ V j · ∇ V j = −∇p j + n j q j E + Vj × B + K j

n jm j (8.7)
∂t
where we have now retained the separate electric and magnetic forces, and introduced an interaction
force K j which represents momentum added to species “j” from the other species in the system. This
momentum transfer mimics collisions and other kinetic effects not included directly in the multi-fluid
description. Since the total momentum of the plasma must be conserved, one species’ gain must be
another’s loss, so that ∑ j K j = 0.
We could also adopt an ad hoc polytropic approximation here, as we did in Chapter 4. However, we
might need to incorporate energy transfers amongst the different species, and so we instead derive an
energy equation based on the thermodynamics of a fluid element. The first law of thermodynamics tells
us
dU = dW + dQ (8.8)
where U is the internal (i.e., thermal) energy, dW is the external work done on the element (by the
pressure forces on it), and dQ is the heat added to the element. For an element of volume V , there is
1
2 kb T j of thermal energy in each degree of freedom. If we assume that the species is monatomic then
there are only the 3 translational degrees of freedom, and so U = 32 pV . Now the net work (force times
distance) done by the external pressure is simply dW = −p d V . Thus the first law becomes
! !
3 3 3
d pV ≡ V d p + p d V = −p d V + dQ (8.9)
2 2 2
Here d denotes a change in a quantity following the fluid element, and so should be associated with
the convective derivative ∂/∂t + V j · ∇ similar to (4.5). [This is not strictly true for the dQ term since,
thermodynamically, dQ is not a real differential, but that will not concern us here.] Noting that the
continuity equation (8.6) can be written d(n j V ) = 0, (following an element, the number of particles it
contains doesn’t change) dividing by V , collecting terms, and using (8.6) leads to
∂
! !
3 5
+Vj ·∇ p j = − p j∇ · V j + H j (8.10)
∂t 2 2
where H j represents the heat added per unit volume to the jth species from all the others. It is straight-
forward to show that, in the absence of heat addition, Equation 8.10 reduces to (4.3) with γ = 5/3.
We need only add Maxwell’s equations to complete the system. In particular, we need the induction
equation
∂B
= −∇ × E (8.11)
∂t
together with Poisson’s equation
ρq 1
∇·E =
ε0
=
ε0
∑ n jq j (8.12)
j
and Ampere’s Law including the displacement current

1 ∂E
∇ × B = µ0 j +
c2 ∂t
1 ∂E
= µ0 ∑ n j q j V j + (8.13)
j c2 ∂t
8.2. GENERALISED OHM’S LAW 103
Notice that while each species has its own set of equations, they are coupled through Maxwell’s equa-
tions which require the total charge and current densities.
8.2 Generalised Ohm’s Law

The multi-fluid description allows us to re-examine the conducting properties of a plasma since we can
now follow the charge carriers separately. This improves on Equation 4.9 which we wrote down as a
generic Ohm’s Law for a conductor.
We consider a two-fluid system with electrons (“e”) and ions (“i”). For simplicity we assume the
ions are singly ionized. Then we multiply the equation of motion (8.7) of each species by q j /m j to reach
∂
!
e ni e2 e
ni e + Vi · ∇ Vi = − ∇pi + (E + Vi × B) + Ki (8.14)
∂t mi mi mi
∂
!
e ne e2 e
−ne e + Ve · ∇ Ve = + ∇pe + (E + Ve × B) − Ke (8.15)
∂t me me me
We now add these two equations, noting that Ke = −Ki ≡ −K, that ne ≈ ni = n on scales larger than a
Debye length, and that the electric current j ≡ ni eVi − ne eVe . This results in
∂
! ! !
j pi pe 1 1
ne + ne (Vi · ∇Vi − Ve · ∇Ve ) = −e∇ − + ne2 + [E + Ve × B] +
∂t ne mi me mi me
!
e 1 1
+ j × B + eK + (8.16)
mi mi me
The quantity K is the momentum transferred to the ions from the electrons. If this transfer is due
to collisions (or any process which tends to equilibrate the species’ motion), then we expect such a
frictional force to be proportional to the relative motion. That is, we expect K ∝ Ve − Vi ∝ −j so that
we may write, say, K = −nej/σ where σ will turn out to be the conductivity of the plasma. Multiplying
(8.16) by µ/ne2 where µ ≡ mi me /(mi + me ) ≈ me and re-arranging leads to
j
E + Ve × B = −
σ

1 µ µ
− ∇ pe − pi −
ne me mi
Generalised Ohm’s Law µ (8.17)
− j × B+
nemi !
∂ j µ
+µ 2 + (Vi · ∇Vi − Ve · ∇Ve )
∂t ne e
The top line is just the MHD Ohm’s Law (4.9) apart from the explicit dependence on the electron
velocity. This dependence reveals that it is the electrons which are more mobile; the electron frame is the
rest frame of the “conductor.” The next line corresponds to “thermo-electric” fields. Since µ ≈ me mi
we can see that only the electron contribution is likely to be significant. Such electric fields arise because
pressure gradients drive the lighter electrons, leading to a charge separation electric field which prevents
them from running away from the ions.
The j × B term is known as the Hall term due to the presence of the magnetic field. The associated
electric field is perpendicular both to the magnetic field and the electric current. The final set of terms
in (8.17) come from the inertial portions of the equations of motion. They are typically small unless
there are rapid temporal variations or large currents. Hall and inertial effects can be extremely important
on the microscopic scale, however. For example, the initiation and structure of reconnecting plasmas is
controlled by these terms.
8.3 Waves in Cold Two-fluid Plasmas

The multi-fluid formulation is an excellent starting point for exploring the rich wave properties of multi-
species plasmas. In particular, we can relax the MHD assumption of large length scales (L rL , λD )
and long timescales (τ 1/Ωc , 1/ω pe ) although we still cannot include all the kinetic effects.
We proceed in identical fashion to that in Section 4.6.1 by linearising and Fourier analysing the gov-
erning equations. In this case these equations are (8.6), (8.7), (8.10), together with Maxwell’s Equations
(8.11), (8.12),and (8.13). Here, we allow for the species to have relative drifts, so that V j = V j0 + V j1
but we assume that all species are cold, i.e., p j ≡ 0.
This leads to the following set:
(−ω + k · V j0 )δn j + (k · δV j ) n j0 = 0 (8.18)
im j n j0 (−ω + k · V j0 ) δV j = n j0 q j (δE + δV j × B0 + V j0 × δB) +
+δn j q j (V j0 × B0 ) (8.19)
k × δE = ωδB (8.20)
δn j q j
ik · δE = ∑ (8.21)
j ε0
iωδE
ik × δB = µ0 ∑ q j (δn j V j0 + n j0 δV j ) − (8.22)
j c2
Note that ∇ · δB = 0 is preserved by (8.20) if it is satisfied initially, so we have not included that equa-
tion. Indeed, dotting (8.22) with k and using the continuity result (8.18) reveals that Poisson’s equation
is also redundant (in effect, continuity of mass implies continuity of charge). Nonetheless, for some
applications Poisson’s equation is more convenient, so we shall keep both here.
Plasma texts (e.g., that by Boyd and Sanderson) systematically explore the variety of waves which
can exist in even a relatively simple, two-fluid (ions and electrons) plasma. We shall content ourselves
here with a couple of examples which are not accessible to MHD treatment.
8.3.1 Two-stream Instability

Let us start with a plasma composed of two equal, counterstreaming (electron) beams in the absence of
a magnetic field. Consider for simplicity a 1-D problem in which k, the velocities (both background and
fluctuations), and the fluctuating electric field are all co-linear. Thus V j0 = ±V0 ≡ ±V0 k̂ and n j0 = n0 .
There will be a background of a third species (ions, to provide overall charge neutrality) which we will
assume is dynamically unimportant. The governing equations reduce in this case to:
− (ω ∓ kV0 )δn j + kδV j n0 = 0 (8.23)

−im(ω ∓ kV0 )δV j = qδE (8.24)
q
ikδE =
ε0
∑ δn j (8.25)
j=1,2
8.3. WAVES IN COLD TWO-FLUID PLASMAS 105
where the upper sign corresponds to j = 1 and the lower to j = 2. Solving for δn j yields
" #
n0 kδV j n0 k − qδE
δn j = =
ω ∓ kV0 ω ∓ kV0 im (ω ∓ kV0 )
in0 q k
= δE (8.26)
m (ω ∓ kV0 )2
Substituting this result into (8.25) yields an equation in δE:

" #
q in0 q 1 1
ikδE = kδE + (8.27)
ε0 m (ω − kV0 )2 (ω + kV0 )2
or " #
1 1
δE = ω2p + δE (8.28)
(ω − kV0 )2 (ω + kV0 )2
where ω2p ≡ n0 q2 /mε0 is the square of the plasma frequency of each beam. Non-trivial δE thus requires
" #
1 1
1 = ωp 2
+ (8.29)
(ω − kV0 )2 (ω + kV0 )2
which is the desired dispersion relation. A couple of lines of algebra yields
ω4 − 2ω2 k2V02 + k2V02 k2V02 − 2ω2p = 0

(8.30)
which is a quadratic equation in ω2 . Expanding the solution for small k (i.e., k2V02 /ω2p 1 yields two
approximate roots
ω2 ≈ 2ω2p + 3(kV0 )2 (8.31)
which is just the dispersion relation for plasma oscillations modified slightly by the beam relative drifts
and
Two Stream Instability ω2 ≈ − (kV0 )2 (8.32)
This represents two modes, ω = +ikV0 and ω = −ikV0 . Recalling that the fluctations vary as ei(k·x−ωt)
we see that one of these modes decays as e−(kV0t) while the other grows as e+(kV0t) . Thus the initial
configuration is unstable, the (presumed small) fluctuations grow until a nonlinear regime is reached.
Typically, the nonlinear stage of an instability tends to destroy or disrupt the “free energy” which drives
the instability in the first place. In the present case this would be through a deceleration of the relative
beam motion, turning that energy into electric energy and ultimately heat.
In general, plasmas don’t like beams, currents, etc., and are usually unstable to waves of some kind
for any significant beam or current features. The nonlinear aspects of such instabilities scatters plasma
particles, reducing the beam/current, and thus subsuming the role of collisions. The consequences of
such scattering on the current carried by the plasma is termed “anomolous resistivity.”
8.3.2 Electron Whistler Waves

Let us now explore one of the “natural” plasma modes, known as a whistler wave. Whistlers were
important in the history of understanding the Earth’s magnetic field and atmosphere. We shall see that
they are dispersive; higher frequency waves travel faster and thus arrive first, giving the characteristic
Figure 8.1: Dynamic spectrum of frequency vs. time showing several sweeping arcs characteristic of
whistler waves in the Earth’s atmosphere. [From Burgess and Inan, 1993]
falling tone (which you can hear on a shortwave radio) from which they are named. These features can
be seen in the observations shown in Figure 8.1. Moreover, whistlers can travel from one hemisphere
along the Earth’s dipole field to the other, and thus travel through the “magnetosphere.” Knowing their
dispersion properties allowed ionospheric physicists to determine the density and other parameters from
the waves received on the ground.
We return to the basic cold multi-fluid wave equations (8.18)-(8.22). In this case, we consider only
one species, electrons, and again assume that there is a background of ions which, due to their larger
mass, are dynamically unimportant. We assume that there is no drift, i.e., V j0 = 0 and that the wave-
vector k is aligned with the equilibrium magnetic field B0 . Whistlers turn out to be transverse, so we
shall save some algebra by assuming that there is no longitudinal fluctuation, i.e., k · δV = 0. Then we
see immediately that there are no density fluctuations, and hence there is no longitudinal electric field,
so that we can dispense with both the continuity and Poisson’s equations. Eliminating δB using (8.20)
leaves:
− imωδV = qδE + qδV × B0 (8.33)
i iω
k × (k × δE) = µ0 qn0 δV − 2 δE (8.34)
ω c
Note that k×(k × δE) = k (k · δE)−k2 δE = −k2 δE. We shall now solve (8.33) for δV and substitute the
result into (8.34). To do this, we introduce a coordinate system with B0 = B0 ẑ. The x and y components
of (8.33) can be written
δVx δEx

−iω −Ω q
= (8.35)
Ω −iω δVy m δEy
where Ω ≡ qB0 /m is the cyclotron frequency. Inverting the matrix of coefficients leads to
δVx − q/m −iω Ω δEx

= 2 (8.36)
δVy ω − Ω2 −Ω −iω δEy
8.3. WAVES IN COLD TWO-FLUID PLASMAS 107
Substituting this into (8.34) and a bit of tidying results in
δEx ω2p ω δEx

−ω −iΩ
k c −ω
2 2 2

= (8.37)
δEy ω2 − Ω2 iΩ −ω δEy
which can be cast into a homogeneous linear system of the form
2 c2 − ω2 + ω p ω + iω2p ωΩ
2 2
 
k
ω2 − Ω2 ω2 − Ω2  δEx

0

A · δE ≡  = (8.38)

− iω2p ωΩ ω2 ω2  δEy 0

k 2 c2 − ω2 + 2 p 2

ω −Ω
2 2
ω −Ω
where A is related to the dieletric tensor of the medium. Non-trivial solutions require det A ≡ 0 from
which the dispersion relation can be found. Indeed, the first step can be seen practically by inspection to
yield
k 2 c2 ω2 ω2 ωΩ
− + =± 2 (8.39)
ωp 2
ωp ω − Ω
2 2 2
ω − Ω2
We now evaluate this in the limit that ω2 ω2p . Ignoring then the second term and multiplying by
(ω2 − Ω2 ) gives a simple quadratic
!
2 2
k c k 2 c2
ω2 1 + 2 ∓ ωΩ − 2 Ω2 = 0 (8.40)
ωp ωp
Applying the quadratic formula (note that the ∓ sign in the above is independent of that in the quadratic
formula), and expanding the result in the long wavelength regime (k2 c2 /ω2p 1) yields two different
modes (each of which can propagate in either directions so that there are four roots altogether). One is a
cylcotron resonance (ω = ±Ω) while the other is
k2 c2 k2 vA2 mi
Cold Whistler Mode ω = ±Ω ≡ (8.41)
ω2p Ω me
Notice that these whistler waves have ω ∝ k2 , and so the phase speed ω/k increases with k; shorter
wavelength, higher frequency whistler waves travel faster. Moreover, the group velocity ∂ω/∂k is faster
than this phase velocity (by a factor of 2) and also increases with k. Thus the energy/information carried
by a whistler likewise doesn’t travel at a single speed. Typically, ionospheric whistlers are excited by
lightning and other transient phenomena. This single spike, which is comprised of a range of frequencies,
then “disperses” as it travels along the magnetic field.
8.3.3 Other Waves

The derivation of the whistler dispersion relation reveals some of the complexity and richness of waves
in plasmas. The more general case, in terms of real multi-fluids (i.e., including ions which participate
in the wave motion), is a fascinating subject which is complicated but still manageable. The results
can be collected, in the case of cold plasmas at least (i.e., those for which all the thermal pressures
are ignorable), into a single diagram, known as the Clemmow-Mullaly-Allis or CMA diagram. This is
shown in Figure 8.2. It represents the phase speeds of the various modes under various frequency and
plasma regimes in the same way that the Friedrich’s Diagram of Figure 4.8 shows the phase fronts of the
MHD waves in an MHD “pond.” Indeed, the fast and Alfvén portions of the MHD Friedrichs Diagram
can be recognised in the top right portion of Figure 8.2.
Further progress on waves and instabilities involving plasma populations requires abandoning the
fluid treatments and commencing with kinetic, Vlasov theory. This is a fascinating subject, with inter-
esting mathematics and physics at each junction, but is beyond the scope of the present course.
8.4 Exercises
1. Starting from the first law of thermodynamics, fill in the derivation of the internal energy equation
(8.10) and verify the claim that it reduces to a polytrope with γ = 5/3 in the absence of heat
addition.
2. If the momentum transfer between ions and electrons can be represented by a simple stream-
ing friction, so that Ki = νc me n(Ve − Vi ), follow the derivation of (8.16) and (8.17) to find an
expression for the conductivity σ in terms of the collision frequency νc and other constants or
parameters.
3. Show that in the general cold multi-stream problem the dispersion relation for purely electric
waves in the absence of a magnetic field is
ω2p j
∑ (ω − k · V 2
=1 (8.42)
j j0 )
4. Fill in the steps from (8.33) leading to the whistler dispersion relation (8.41).
8.4. EXERCISES 109
Figure 8.2: Clemmow-Mullaly-Allis (CMA) diagram for waves in a two-fluid, cold plasma. The verti-
cal axis is Ωe /ω which reveals the magnetic field dependence of the modes, while the horizontal axis
is ω2pe /ω2 , which shows the density dependence. For convenience, an unrealistically small mass ra-
tio mi /me has been chosen to compute this diagram. The space is divided into various regions within
which the wave dispersion properties don’t change. Within each region are shown figures of constant
phase for the different modes propagating in different directions with respect to the background mag-
netic field. These can be thought of as the ripples made by dropping pebbles into a plasma pond, and
extends the Friedrichs Diagram shown in Figure 4.8 to higher frequencies and other regimes. [From
www.physics.auburn.edu/ swanson/chap2.pdf]
Appendix A
Useful Mathematical Results and Physical

Constants
A.1 Physical Constants

Quantity symbol Value
Proton Mass mp 1.6726 × 10−27 kg
Electron Mass me 9.1095 × 10−31 kg
Electron-to-proton mass ratio m p /me 1836.2
Speed of light in vacuum c 2.9979 × 108 m/s
Gravitational Constant G 6.672 × 10−11 N m2 kg−2
Boltzmann constant kb 1.3807 × 10−23 J/K
Electronic charge e 1.6022 × 10−19 C
Permittivity of free space ε0 8.8542 × 10−12 F/m
Permeability of free space µ0 4π × 10−7 H/m
Mass of the Earth ME or M⊕ 5.972 × 1024 kg
Radius of the Earth Re or RE 6.3712 × 106 m
Magnetic Dipole moment of Earth 8 × 1022 A m2 = 0.3 × 10−4 T R3E
Surface Magnetic Field of Earth 3 × 10−5 T
Mass of the Sun M 1.99 × 1030 kg
Radius of the Sun R 6.96 × 108 m
Sun-Earth distance AU 1.50 × 1011 m = 215R
A.2 Vector Identities

A·B×C = B·C×A = C·A×B (A.1)
A × (B × C) = (C × B) × A = B (A · C) − C(A · B) (A.2)
∇( f g) = f ∇g + g∇ f (A.3)
∇· fA = f∇·A+A·∇f (A.4)
∇ × ( f A) = f ∇ × A + ∇ f × A (A.5)
111
112 APPENDIX A. USEFUL MATHEMATICAL RESULTS AND PHYSICAL CONSTANTS
∇ · (A × B) = B · ∇ × A − A · ∇ × B (A.6)
∇ × (A × B) = A(∇ · B) − B(∇ · A) + (B · ∇) A − (A · ∇) B (A.7)
A × (∇ × B) = (∇B) · A − (A · ∇) B (A.8)
In particular, slightly re-arranged and specialised
(∇ × B) × B = B · ∇B − ∇B2 /2 (A.9)
∇ (A · B) = A × (∇ × B) + B × (∇ × A) + (A · ∇) B + (B · ∇) A (A.10)
∇×∇f = 0 (A.11)
∇·∇×A = 0 (A.12)
∇ × (∇ × A) = ∇ (∇ · A) − ∇2 A (A.13)
A.3 Vector Operators in Various Coordinate Systems

A.3.1 Cartesian Coordinates
In a Cartesian System (x, y, z)
∂f ∂f ∂f
!
∇f = , , (A.14)
∂x ∂y ∂z
∂f ∂f ∂f
∇·A = + + (A.15)
∂x ∂y ∂z
∂Az ∂Ay ∂Ax ∂Az ∂Ay ∂Ax

!
∇×A = − , − , − (A.16)
∂y ∂z ∂z ∂x ∂x ∂y
∂2 f ∂2 f ∂2 f
∇2 f = + + (A.17)
∂x2 ∂y2 ∂z2
A.3.2 Cylindrical Coordinates

In a Cylindrical Polar Coordinate system (r, φ, z)
∂f 1 ∂f ∂f
!
∇f = , , (A.18)
∂r r ∂φ ∂z
1 ∂ 1 ∂Aφ ∂Az
∇·A = (rAr ) + + (A.19)
r ∂r r ∂φ ∂z
A.3. VECTOR OPERATORS IN VARIOUS COORDINATE SYSTEMS 113
1 ∂Az
∂Aφ ∂Ar ∂Az 1 ∂ 1 ∂Ar
!
∇×A = − , − , rAφ − (A.20)
r ∂φ ∂z ∂z ∂r r ∂r r ∂φ
1 ∂ ∂f 1 ∂2 f ∂2 f
!
∇2 f = r + 2 2+ 2 (A.21)
r ∂r ∂r r ∂φ ∂z
A.3.3 Spherical Polar Coordinates

In a spherical polar coordinate system (r, θ, φ)
∂f 1 ∂f 1 ∂f
!
∇f = , , (A.22)
∂r r ∂θ r sin θ ∂φ
1 ∂ 1 ∂ 1 ∂Aφ
∇·A = r2 Ar + (Aθ sin θ) +

(A.23)
r ∂r2
r sin θ ∂θ r sin θ ∂φ
∂ ∂Aθ ∂Ar 1 ∂1 ∂ 1 ∂Ar

!
1 1 1
∇×A = Aφ sin θ −

, − rAφ , (rAθ ) − (A.24)
r sin θ ∂θ r sin θ ∂φ r sin θ ∂φ r ∂r r ∂r r ∂θ
1 ∂ 2∂f ∂ ∂f ∂2 f
! !
1 1
∇ f=
2
r + 2 sin θ + 2 2 (A.25)
r2 ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂φ2
∂Br Aθ ∂Br Aφ ∂Br Aθ Bθ + Aφ Bφ

(A · ∇) B = Ar + + − r̂
∂r r ∂θ r sin θ ∂φ r
∂Bθ Aθ ∂Bθ Aφ ∂Bθ Aφ Bφ

Aθ Br
+ + Ar + + − cot θ θ̂ (A.26)
r ∂r r ∂θ r sin θ ∂φ r
∂Bφ Aθ ∂Bφ Aφ ∂Bφ

Aφ Br Aφ Bθ
+ + cot θ + Ar + + φ̂
r r ∂r r ∂θ r sin θ ∂φ
114 APPENDIX A. USEFUL MATHEMATICAL RESULTS AND PHYSICAL CONSTANTS
Appendix B
The Mathematics of Waves
B.1 Fourier Analysis

Many branches of physics involve the study of wave properties of the media in question. These waves
illustrate the way the various restoring forces in the medium manifest themselves in the dynamics. In
the case of linearised equations of motion, the resulting perturbations can always be broken down into
any convenient summation of linear modes. Particularly convenient and common is the use of Fourier
techniques to represent the disturbance as a linear superposition of harmonic plane waves.
In the case of a simple function of time g(t), Fourier’s theorem enables us to re-write this in terms
of a Fourier Series as:
∞
a0
g(t) = + ∑ an cos 2nπt/τ + bn sin 2nπt/τ (B.1)
2 n=1
where Z +τ/2
an 2 cos 2nπt/τ
= g(t) dt (B.2)
bn τ −τ/2 sin2nπt/τ
where τ is time interval under consideration. The an ’s and bn ’s give the amplitude of the nth harmonic
of the fundamental period τ, i.e., at a frequency fn = n/τ.
If we let the time interval τ get larger and larger, the interval between successive modes gets smaller
and smaller. In the limit as τ → ∞ the discrete sum in Equation (B.1) gets replaced by an integral which
defines what is known as the Fourier transform
Z +∞
1
g(t) = √ g̃(ω)e−iωt dω (B.3)
2π −∞
where now the Fourier transform g̃(ω) is

Z +∞
1
g̃(ω) = √ g(t)eiωt dt (B.4)
2π −∞
Now g̃(ω) provides the (differential) amplitude of the disturbance at an angular frequency ω ≡ 2π f .
Note that in this continuous frequency approach for the decomposition of g(t) there is a certain degree
of arbitrariness in the normalisation of g̃ and other treatments may write the above with the factor 2π
distributed differently between the transform and its inverse. Additionally, traditional mathematical
treatments use the opposite sign convention for ω in the exponential functions. The above is used here
to reinforce the physics convention that propagating waves have a time dependence ∝ e−iωt .
The power of Fourier analysis lies in the fact that it is complete, in the sense that any (integrable)
function g(t) can be decomposed into its Fourier components in a unique way, and the resulting Fourier
series/transform converges to g(t) everywhere except at places where g(t) is discontinuous. Moreover,
115
116 APPENDIX B. THE MATHEMATICS OF WAVES
differentiating g(t) with respect to t merely multiplies the integrand in Equation (B.3) by −iω, so that
time derivatives become algebraic multiplications in the frequency domain. Thus, for example, the
equation
dg
= h(t) (B.5)
dt
can be Fourier transformed to give:
Z +∞ Z +∞
−iωg̃e−iωt dω = h̃e−iωt dω (B.6)
−∞ −∞
But since the complex exponential functions are complete and orthogonal, this requires simply
− iωg̃ = h̃ (B.7)
thereby enabling the original differential equation to be solved algebraically for each Fourier component.
B.2 Waves in Space and Time: 1D

B.2.1 Phase Speeds
Fourier techniques can be applied to good advantage in the study of waves which propagate in space as
well as oscillating in time. In one dimension we have:
Z +∞ Z +∞
g(x,t) = g̃(k, ω)ei(kx−ωt) dω dk (B.8)
−∞ −∞
where k = 2π/λ is the wavenumber of the wave, whose wavelength (distance between wave crests) is λ.
It is customary to define the k-dependence with the opposite sign convention to that used for ω so that
points of constant wave phase have
phase = kx − ωt = constant (B.9)
For example, points at phase = 0 satisfy kx − ωt = 0, or
ω
x− t ≡ x − v pht = 0 (B.10)
k
where obviously
ω
Phase Speed v ph = (B.11)
k
is the wave phase speed. In terms of the wavelength λ, period T , and frequency f we have
ω 2π f λ
v ph = = = λf = (B.12)
k 2π/λ T
which simply says that a wave travels one wavelength in one period.
B.2. WAVES IN SPACE AND TIME: 1D 117
B.2.2 Group Speed

For simple waves, the phase speed also represents the speed at which information and/or energy are
transported by the waves. This is the case for “non-dispersive” media, in which v ph is a constant (in-
dependent of the wavelength or frequency). In dispersive media, however, the phase speed can be very
different from the speed of information transport. Since we are considering cases where ω/k are not
constant, there will be some relationship between the wave angular frequency and wavenumber which
is known as the dispersion relation. For convenience we denote this as
ω = ω(k) (B.13)
although there may be occasions where the dispersion in time is more relevant, and one might consider
k = k(ω) as the dependent variable.
We now construct a wavepacket which is localised in space and seek to determine not only the propa-
gation of the points of constant phase, but also the displacement of the bulk of the wave amplitude. Such
a packet is constructed from a set of wavenumbers within an interval ∆k of some central wavenumber
k0 , so that Z +∞
g(x,t) = A(k)ei(kx−ω(k)t) dk (B.14)
−∞
where A(k) is a function peaked at k0 , and we have used the dispersion relation to perform the integration
over ω, which merely picks out the ω satisfying ω = ω(k). Notice that at t = 0 the envelope of the wave
packet can be found by re-writing the above as
Z +∞
g(x,t = 0) = eik0 x A(k)ei(k−k0 )x dk (B.15)
−∞
The pure exponential outside the integral is the “carrier” wave, while the integral gives the shape of the
envolope which modulates these oscillations. We take advantage of the form of A(k) by expanding ω(k)
about k0 , so that

dω
kx − ω(k)t ≈ (k − k0 )x + k0 x − ω(k0 )t − (k − k0 )t (B.16)
dk k0
!
dω
= (k0 x − ω0t) + (k − k0 ) x − t (B.17)
dk
where ω0 ≡ ω(k0 ). We define the group speed by
dω
Group Speed vg = (B.18)
dk
Thus our disturbance obeys

Z +∞
g(x,t) = e i(k0 x−ω0 t)
A(k)ei(k−k0 )(x−vgt) dk (B.19)
−∞
Comparing this with Equation (B.15) shows that the phase oscillations of the carrier propagate as ex-
pected at the phase speed ω0 /k0 from the part in front of the integral. This is a totally un-localised wave.
Inside the integral is the envelope of these oscillations. This envelope is just that found in (B.15) but
evaluated at x − vgt instead of x; that is, the envelope moves at a speed vg .
Figure B.1: Dispersion curves (solid) for plasma oscillations (ω = ω pe = const) and whistler waves
showing the corresponding phase and group speeds at specific (ω, k) points (filled circles).
For simple waves, such as light or sound waves in an ordinary gas, the dispersion relation is of the
form
ω = kc (B.20)
and both the phase and group speeds are equal to c. In more complicated media, this is no longer the
case. Consider the case of plasma oscillations introduced in Section 2.5.1 for which
ω(k) = ω pe (B.21)
Since ω pe is a constant, the phase speed is simply v ph = ω pe /k and increases with decreasing k. As k → 0
this phase speed can exceed the speed of light! However, this represents nothing more than the phasing
between oscillations in neighbouring plasma regions, as though there were a set of pendulums suspended
along a line. The speed with which a displacement appears to move along the line is determined only by
the initial phases of the individual pendulums; no information nor energy is transferred from one to the
other.
dω
By contrast, for such plasma oscillations, the group speed is vg = dkpe = 0. No information is
transferred through the medium. A local set of oscillators continue to oscillate in place and do not
send any signals beyond their locations. These properties are often illustrated by means of a dispersion
diagram ω vs. k. The slope of the line from the origin to the relevant point on the dispersion curve ω(k)
gives the phase speed v ph = ω/k while the slope of tangent to the curve at that point is the group speed.
These are illustrated in Figure B.1.
B.2. WAVES IN SPACE AND TIME: 1D 119
B.2.3 Waves in 3D
The analysis in the preceding sections is readily extended to the analysis of plane waves in three dimen-
sions by the specification of three wavenumbers along the three cartesian directions, i.e., k = (kx , ky , kz ).
Now our archetypal wave has disturbances which propagate according to
ei(k·x−ωt) (B.22)
where x = (x, y, z). The vector k is known as the wavevector. Points which have the same value of k · x
at a given instant have the same phase, so k is directed perpendicular to the wave phase fronts and has a
magnitude equal to 2π over the distance between successive fronts, i.e., |k| = 2π/λ. In these cases the
phase velocity vph is now a vector. Only the component along k is unique, since adding any velocity
along the planar wave front yields a phase pattern which is indistinguishable from any other. Thus it is
customary to think of vph as being directed along k so that
ω k
Phase Velocity vph = (B.23)
|k| |k|
while the group velocity is now
∂ω
Group Velocity vg = (B.24)
∂k
which is uniquely defined.

For example, an Alfvén wave propagating in a uniform medium in which the background magnetic
field lies along the ẑ direction has a dispersion relation given by
ω(k) = kz vA (B.25)
where vA is a constant (see Equation (4.59)). For such waves, the phase velocity, as depicted in the
Friedrichs diagram shown in Figure 4.8, is
k kz k
vph = ω = vA (B.26)
|k|2 |k|2
The phase velocity goes to zero at perpendicular propagation, and is less than the Alfvén speed unless k
is aligned with the magnetic field.
On the other hand, the group velocity in this case is
∂ω
vg = (B.27)
∂k
∂kz vA ∂kz vA ∂kz vA

= , , (B.28)
∂kx ∂ky ∂kz
= (0, 0, vA ) (B.29)
showing that, regardless of the direction of k, the energy transport in an Alfvén wave is always purely
along the magnetic field at the Alfvén speed.
Index
β, 49 distribution function, 12
Alfvén draping of field lines, 40
radius, 67 drift, 26
speed, 46 E × B, 23
00Contents, 2 curvature, 27
gradient, 28
adiabatic invariant, 32 dynamo, 50
anomolous resistivity, 105 α − ω, 53
Archimedean spiral, 66 Cowling’s Theorem, 51
kinematic, 51
books, 9 mean field, 52
boundary layer
cell model, 41 electric field
butterfly diagram, 53 motional, 36
exact description, 12
cell model, 41
CIR, 67 fast magnetosonic waves, 47
CME, 67 flux tube, 40, 49
collisions force
Coulomb, 18 Lorentz, 13
frequency, 18 force free, 49
conservation forces
charge, 13 MHD, 41
constants,physical, 111 Fourier
continuity, 102 series, 115
Coplanarity, 94 transform, 115
coronal heating, 68 frequency
Coronal Mass Ejections, 67 cyclotron, 22
Cowling’s Theorem, 51 Friedrichs Diagram
critical radius, 63 MHD, 48
curvature drift, 27
gradient drift, 28
de Hoffmann-Teller Frame, 97 group velocity, 119
Debye guiding centre, 29
length, 17 Hale’s Law, 55
number, 17 Hall term, 104
sheilding, 17
diffusion induction equation, 37
magnetic, 50 instabilities
discontinuities MHD, 50
MHD, 90 two stream, 104
solar wind, 68 interplanetary magnetic field, 66
121
122 INDEX
Larmor radius, 22 guiding centre, 29

Lorentz force, 21 Larmor radius, 22
Lorentz force, 21
Mach Number magnetic mirroring, 32
Alfvén, 93 magnetic moment, 29
sonic, 93 pitch angle, 29
magnetic relativistic, 33
diffusion, 50 uniform B, 21
annihilation, 71 Petschek reconnection, 80
mirror, 32 phase speed, 116
moment, 29 pitch angle, 29
invariance, 32 plasma
pressure, 43 astrophysical examples, 11
reconnection(see reconnection), 71 beta, 49
tension, 43 definition, 9, 11
magnetohydrodynamics, 13 plasma frequency, 15
magnetohydrodynamics, (see MHD), 35 plasma oscillations
magnetosphere, 75 charge neutrality, 15
mathematical properties, 115 two fluid, 105
Maxwell’s Equations, 13 Poynting Flux, 79
MHD, 13, 35 pressure
discontinuities, 90 magnetic, 43
dynamo, 50
equations, 14 Rankine-Hugoniot relations, 87
equilibria, 49 reconnection, 71
current free, 49 diffusion region, 74
flux tube, 49 magnetopause data, 82
force free, 49 magnetospheric, 75
forces, 41 Petschek, 80
ideal, 37 Petschek rate, 82
induction, 37 Sweet-Parker, 78
instabilities, 50 resistivity
momentum anomolous, 105
conservation, 89 Reynold’s Number, Magnetic, 37
total energy
shocks, 85
conservation, 89
θBn , 96
waves, 43
classification, 90
multi-fluid models, 101
collisionless, 86
Ohm’s Law, 36 conservation relations, 87
generalised, 103 Coplanarity Theorem, 94
Hall term, 104 de Hoffmann-Teller Frame, 97
parallel, 91, 96
Parker spiral, 66 perpendicular, 92, 96
particle motion quasi-parallel, 96
adabatic invariant, 32 quasi-perpendicular, 96
curvature drift, 27 Rankine-Hugoniot relations, 87
cyclotron frequency, 22 reflected particles, 96
drifts, 26 solar wind, 68
gradient drift, 28 subcritical, 97
INDEX 123
supercritical, 97
solar flare, 75
solar wind, 59
Coronal Mass Ejections, 67
Corotating Interaction Regions, 67
critical radius, 63
discoutinuities, 68
magnetic field, 66
parameter values, 60
Parker spiral, 66
Parker’s Equation, 63
sectors, 67
shocks, 68
simple model, 61
static solution, 61
streams, 67
speed
group, 117
phase, 116
tension
magnetic, 43
transformation
E,B Lorentz, 14
two fluid models, 101
two stream instability, 104
vector
identities, 111
operators, 112
Vlasov equation, 14
waves, 115
Alfvén, 46
fast and slow magnetosonic, 48
fast magnetosonic, 47
group speed, 117
group velocity, 119
MHD, 43
Friedrichs Diagram, 48
phase speed, 116
phase velocity, 119
sound, 45
wavevector, 119
whistler, 105
wavevector, 119
whistler waves, 105

AP Master

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

AP Master

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AP Master

Uploaded by

Copyright:

Available Formats

Astrophysical Plasmas

Steven J. Schwartz, Christopher J. Owen1 , and David Burgess

19 November 2002; last revision 4 January 2004

1 Introduction and Overview 9

2 Some Basic Concepts 11

3 Individual Particle Motion 21

5 The Solar Wind 59

7 Shocks and Discontinuities 85

8 Multi-Fluid Models 101

A Useful Mathematical Results and Physical Constants 111

B The Mathematics of Waves 115

Introduction and Overview

• R.O. Dendy, Plasma Dynamics, Oxford, 1990. Good. Available in paperback.

• Nicholson, Introduction to Plasma Theory, Wiley.

• Parker, Cosmical Magnetic Fields, Clarendon Press, 1979. Advanced.

Some Basic Concepts

2.1 What is a Plasma?

2.2 Plasmas in the Universe

• Earth’s (and other planets’) ionosphere (above 60 km) and magnetosphere

• Sun’s and other stars’ atmospheres, and winds

• Cosmic Rays (galactic and extra-galactic energetic particles)

• Jets in active galaxies - radio jets and emission

• Pulsars and their magnetosphere

Figure 2.1: The realm of plasma physics.

• Accretion disks around stars

2.3 Different ways of describing plasmas

2.4 Governing Equations

2.4.1 Maxwell’s Equations

where µ0 ε0 = 1/c2 , ρq is the charge density, and j is the current density.

2.4.2 Conservation of Charge

2.4.3 Lorentz Force

For an extended medium there is a force density of ρq E + j × B

2.4.4 Magnetohydrodynamics (MHD)

where σ is the electrical conductivity.

2.4.5 Kinetic Theory

2.4.6 Frame transformations

2.5 Some Basic Plasma Phenomena

2.5.1 Plasma oscillations

Debye Number ND = n0 λ3D (2.24)

2.5.2 Charge Shielding

2.5.3 Collisions, Mean Free Path, and Collisionless Plasmas

And we can define an effective Coulomb cross-section,

Individual Particle Motion

3.1 Equation of Motion

3.2 Motion in uniform, static magnetic field

3.3 Motion in uniform, static magnetic and electric fields

(Check by substitution and using the vector identities (E × B) × B = (E · B) B − (B · B) E and E · B ≡

3.3.2 Mathematical Treatment

3.4 Particle Drifts

3.4.1 Gravitational Drifts

3.4.2 Curvature Drift

via Equation 3.30.

3.4.3 Gradient Drift

3.5 Useful Parameters for Describing Particle Motion

3.5.2 Guiding Centre

3.5.3 Magnetic Moment

Figure 3.5: Magnetic configuration in which ∇B k B - a “magnetic mirror” geometry.

3.6 Motion in Static, Non-uniform Magnetic Field: ∇B k B